Search the archive.

Currently, the Internet has only seen cyber skirmishes. State actors like North Korea, China, Russia, Iran, and the United States have digitized corporate and governmental espionage, which has a long and rich history. Meanwhile, criminal gangs make the occasional headline with thefts from ATMs and cryptocurrency accounts. Fortunately, there has been very little weaponized use of cybersecurity technology. The most recent example is the American government’s disabling the Iranian military computer network in mid-2019 in retaliation for the Iranians downing a U.S. drone. Unfortunately, the federal government, its contractors, nor private enterprises have as yet developed a sense of MAD: Mutually Assured Destruction. MAD was the unofficial policy — more an understanding, really — between the United States and the Soviet Union during the Cold War. What each knew at the time was that if either side uses nuclear weapons against the other, it could mean the end of the world. Fortunately, nations have held back from using the Internet to shut down the power grids of the United States, or bring airliners crashing to the ground, or launch American missiles against itself. Corporations, however, have not yet developed that level of cyber-awareness about the possible repercussions of hack backs. Unfortunately, in the United States, combatants are fighting purely defensive pitched battles that frustrate them greatly and sometimes cost them dearly. Nevertheless, as cybersecurity tools of mass destruction arrive on the scene, those charged with defending the nation need to closely guarded and wield those weapons with the greatest wisdom and restraint to avoid MAD. Providing the right for anyone to hack back may lead to short-term satisfaction, but long-term devastation in the virtual and real worlds.

Citizen co-production of cybersecurity may also be viewed through the lens of regulatory theory. The introduction to this article referred to the interplay of public and private social control, as noted by Donald Black (1976). To the extent that non-state or informal institutions of social control are functioning effectively, there will be less need for state intervention. When, on the other hand, state institutions of social control are perceived to be operating unsatisfactorily, one may expect non-state actors to seek to compensate. So it is that individual owners of intellectual property engage in self-help when state institutions cannot assure the security of their products (Tusikov 2014). Regulatory pluralism also provides a useful framework for the analysis of cybersecurity co-production. The concept of regulatory pluralism is derived from that of legal pluralism, which is based on the recognition that the law exists alongside a variety of lesser normative orderings. Galanter (1981, p. 20) observed that the legal system is often a secondary rather than a primary locus of regulation. Griffiths (1986, p. 4) referred to legal reality as not monolithic, but rather an “unsystematic collage of inconsistent and overlapping parts.” Julia Black (2001) referred to a “postregulatory world” in which governments no longer monopolize regulation, and later noted the complexity and fragmentation of regulatory systems (Black 2008). Chang (2012b) examined the importance of involving non-state actors (both commercial and non-commercial) as regulators in fostering computer security incident reporting. Other scholars have observed less choreography in regulatory systems. von Hayek (1991) observed that while some orderings are designed and coordinated, others are spontaneous. More recently, Scott (2004) observed that regulatory power can be widely dispersed, and that the state is no longer the main locus of control over social and economic life (if it ever was). Black (2008) speaks of “polycentric” regulatory regimes, which may be ungovernable. Some researchers have proposed a theoretical framework of nodal governance for the policing of cyberspace that conveys the notion that policing needs to be conducted through the collaborative efforts of plural actors (nodal clusters), including criminal investigators, private industry players, and individual internet users (Huey et al. 2013). And with the rise of internet vigilantism (netilantism) and public-private cooperation (co-production) in cybercrime investigation, there is a need for us to understand how private sectors and netizens can help in the co-production of cybersecurity and the risks that may accompany it (Brenner 2007; Chang 2012b; Huey et al. 2013; Nhan et al. 2015; Chang & Poon 2016). Consider the domain of online child exploitation, a type of cybercrime that is the subject of widespread concern (Cohen-Almagor 2013). A high priority for law enforcement agencies in most advanced industrial societies, online child protection activities are also undertaken by a variety of non-state actors and institutions, including parents, teachers, non-government organizations specializing in internet information and education, manufacturers of blocking and filtering software, and self-appointed internet sleuths, operating individually or in concert. As we will see, the efforts of these various players are not systematically choreographed. The activities of some may be coordinated, while others may function independently, for better or worse. While the comprehensive management of internet security may be achievable in more state-centric jurisdictions, this will be less likely in democratic, individualistic societies, where challenges of accountability will be greater.

Despite paying lip service to the problem of corruption for decades, leaders in rich, developed countries have never treated it as more than a second-order foreign policy concern. After all, corruption is hard to measure and easy to brush away with arguments about differing cultural norms and the value of "facilitation payments" in greasing bureaucratic wheels. But lately, it has become harder to deny that corruption lies at the root of many first-order global problems, such as the spread of violent religious extremism or the civil strife and mass casualties witnessed in South Sudan and Syria-not to mention the refugee crises that have followed on their heels. Corruption also plays a major role in the one truly global existential threat: the destruction of the environment When speaking about the causal relationship between corruption and such issues, I'm often asked questions along these lines: "OK, corruption's a bad thing, but is there anything that can be done about it? Are there examples of countries that have pulled themselves back from the brink?" The political scientist Robert Rotberg has surely fielded the same questions countless times during his distinguished career. He has now published a comprehensive and detailed response. His book's answer to the second question is important: some places have indeed dramatically reduced corruption. A few names on that list are familiar success stories, such as Hong Kong and Singapore. Others Botswana, Georgia, Rwanda-might surprise some readers. Rotberg examines these cases, alongside those of both poorer performers and longtime paragons such as Denmark and Finland, in order to figure out what works. His conclusions are scattered throughout the book and then tabulated at the end in a single 14-step program. The book offers authoritative perspectives on a variety of devices that different regimes have applied to the task of fighting corruption. But Rotberg's analysis fails to spell out a reality that his own most fundamental conclusion suggests: corruption is not so much a problem for governments as it is an approach to government, one chosen by far too many rulers today. His suggestions may be helpful to countries that have already undergone some sharp transitions fueled by anticorruption sentiment, such as Brazil, Burkina Faso, Guatemala, South Korea, Tunisia, and Ukraine. The Corruption Cure, however, is less helpful when it comes to hard-boiled kleptocracies, such as Angola and Azerbaijan. Rotberg also downplays the role of developed countries in facilitating such regimes' corrupt practices. And he sidesteps the rather pressing reality of developed countries-including the United States beginning their own unmistakable slides toward kleptocracy. CLEANING HOUSE To fight corruption, a good domestic legal framework is "at least a start," Rotberg writes, as long as it clearly defines illegal behavior and its consequences. In the exemplary case of Singapore, anticorruption laws include "stiff monetary fines and five-year terms of imprisonment for convicted offenders." Civil servants found guilty of corrupt acts can "lose their jobs, their benefits, and their pensions." Apart from punishing acts of corruption after the fact, Singaporean law also does what the U.S. Constitution was at least partly designed to do: prevent corruption before it takes place. (That was the purpose of the Emoluments Clause, which prohibits U.S. officials from receiving gifts from foreign governments-and which is currently the subject of renewed attention owing to three lawsuits charging that Trump has violated it.) In Singapore, Rotberg writes, public servants are prohibited from "borrowing money from or financially obligating themselves to any person with whom they did or could have official dealings," whether or not they have corrupt intentions. Botswana's expansive legislation in this area defines an illegal emolument as "any gift, benefit, loan, or reward; any office, employment or contract, any payments or discharges of obligations or loans; [or] 'any other service,"' Rotberg writes. But too often, Rotberg notes, such laws are just words on paper; what really matters is whether and how they are enforced. Another remedy he examines is anticorruption commissions. In Hong Kong in the 1970s, the Independent Commission Against Corruption (IcAc) made significant inroads against longstanding traditions of illicit gift giving and profiting from official positions. To help the commission carry out its enforcement responsibilities, legislation placed the burden of proof in cases of unexplained wealth on the accused. The commission's independence was bolstered by a generous fixed annual budget and by the fact that the group reported directly to the colonial governor dispatched from the United Kingdom-an official who, Rotberg writes, was considered incorruptible because of his allegiance to London. The ICAC was also subject to oversight by a group of prominent citizens and elected officials. And the commission was charged with corruption prevention and public education, as well as the investigation of wrongdoing. Within a few years of its launch, the ICAC was investigating numerous corruption networks, in which so-called triad gangs worked hand in glove with the police. The commission's prevention department visited government agencies, analyzed their permitting and inspection processes, and recommended improvements. Later, it helped reform the stock exchange and the professional ethics code for lawyers. But such commissions are no panacea, either, Rotberg finds. In many countries, including several in Africa, anticorruption commissions have been worse than ineffective: they have been weaponized to punish opponents of corrupt regimes. In Malawi, for example, Rotberg reports that local observers "believed that the 2004-2006 anticorruption blitz was essentially an exercise in political persecution," since some of its main targets were ranking members of the opposition party. Transparency measures, such as making asset declarations mandatory for public officials, also appear on the list of measures taken by several of Rotberg's "most improved" countries. So do the streamlining of bureaucratic procedures (to remove red tape that might otherwise require a bribe to cut through) and increasing the salaries of civil servants (to reduce the material need to demand or accept bribes). Norms and standards promoted by international institutions can sometimes help, too. In Georgia, Macedonia, and Montenegro, eligibility requirements for EU membership, which all three seek, have catalyzed significant reforms. (Montenegro's efforts so far, however, seem aimed more at checking boxes than at genuinely transforming the way authorities behave.) The list of effective measures also contains several drastic steps, including staff purges at agencies widely seen as thoroughly corrupt. Rotberg reports that when Mikheil Saakashvili came to power in Georgia in 2008, "the new reformers discharged the entire staff of the ministry of education and recruited new employees by competitive examination"; they also sacked 15,000 police officers. In President Paul Kagame's Rwanda, it was "all 503 members of the Rwandan judiciary, from top to bottom," who got the ax, in 2004. And almost immediately after her 2006 election as Liberia's president, Ellen Johnson Sirleaf fired "virtually all" the civil servants who had worked in the prior regime's Ministry of Finance. THE MAFIA STATE The drastic nature of such measures reveals a fundamental weakness in the way The Corruption Cure frames its eponymous problem. Rotberg, like so many authors before him, depicts corruption as an inchoate, corrosive force that seeps into governments that readers might presume are otherwise sound. The metaphor he keeps reaching for is a medical one. Corruption is "an insidious cancer," a "plague" that "infects," "metastasizes," and "cripples." Cure the disease, as the title of the book suggests, and the healed body politic can go out and play. But where does the sickness come from? In explaining how such a malady might take hold, Rotberg resists the temptation to moralize, venturing that corrupt officials may be behaving rationally. "By adopting a conscious strategy of self-enrichment through corrupt behavior, they merely . .. act within the often zero-sum expectations of their class and their condition," he writes. Yet that portrait of widespread but uncoordinated opportunism miscasts the nature of contemporary corruption. Rather than a weakness or a disorder, it is the effective functioning of systems designed to enrich the powerful. Rotberg gestures at this fundamental reality toward the end of the book, when he paraphrases an assessment made by Guatemala's UN-backed anticorruption commission: That country's ruling Patriotic Party "was more a criminal gang than a political party. Its role was to 'rob the state,"' Rotberg writes. In Guatemala, elites "constituted a criminal organization-a kleptocratic conspiracy capable of capturing a national revenue stream, a mafia running a state." This is what corruption looks in at least 60 countries where I have researched the problem: the deliberate operating system of sophisticated networks bent on self-enrichment and remarkably successful at achieving it. For officials in these places, corrupt acts often do not represent rational responses to a permissive environment, as Rotberg would have it; rather, they are a professional requirement. If you are a police officer in Afghanistan or Nigeria, a customs agent in Uzbekistan, or a top administrator in the Honduran environment ministry, you owe your superiors certain things: a cut of your harvest of small bribes, certainly, and perhaps some duly signed and stamped paperwork greenlighting activities that violate regulations. Those who do not perform these allotted tasks are demoted or sidelined-if they're lucky. Sometimes they are shot. It's the old Mafia choice: plata o plomo, "silver or lead." Take the money or take a bullet. These networks come in different forms in different countries. They can be highly structured or fairly diffuse, with varying degrees of internal rivalry and disrupting daily life where they hold sway. Depending on the sources available to them, they capture different revenue streams, including luxury tourism, oil sales, or high-end agricultural exports, such as succulent dates from Tunisia, green beans from Kenya, or the opium whose harvest absorbs much of the labor force in southern Afghanistan each spring. The networks weave together categories that people in developed countries tend to keep separate in their minds: public sector and private sector, black markets and stock markets, professional and personal. Consider the roles played by the Karzai family in post-9/11 Afghanistan, which I had the opportunity to observe at close quarters when I ran a nongovernmental organization established by President Hamid Karzai's older brother Qayum. Karzai served in office for nearly 13 years. Qayum acted as a behind-thescenes power broker, with a stake in a consortium that won millions of dollars in contracts from the U.S. government. Another brother, a self-proclaimed apolitical businessman, owned a cement factory and part of the country's largest private bank, which was later found to operate like a Ponzi scheme. And a third brother served as both a local official and a main facilitator of the region's prodigious opium traffic. In countries such as Azerbaijan, the overlap between the public and private sectors is even more complete, with the ruling family controlling no fewer than 11 banks and sprawling consortia that net the vast bulk of public procurement. In Egypt, the military's control over the economy has vastly expanded under the presidency of Abdel Fattah el-Sisi. The kleptocracy over which Honduran President Juan Orlando Hernindez is striving to gain control remains somewhat more loosely structured: privatesector actors, government officials, and drug traffickers exchange favors and often overlap but maintain a certain degree of separation. In such networks, the role of members who hold public office is to craft laws and regulations and tailor their enforcement in ways that serve the network's aims. In return, they get to loot public coffers or siphon off government revenues; they also get cuts of the bribes extorted at the street level or shares in the companies that their practices benefit. If their activity "destroys developmental prospects" and is "antithetical to economic growth and social betterment" in their country, as Rotberg puts it, that is of no concern whatsoever. Bettering their country's prospects is not their objective. Making money is. SWAMP THING Although Rotberg's disease metaphors elide this reality, his suggested anticorruption program is entirely shaped by it. The first of his 14 steps for a country fighting corruption is that it "seeks, elects, or anoints a transformative political leader." In other words, reforming a severely corrupt country requires nothing short of regime change. In this sense, The Corruption Cure offers a critical warning: once you've toppled your government, make sure you pick a new chief of state on the basis of his or her concrete intentions with respect to corruption. Don't be distracted, for example, by a prospective leader's identity as a political outsider or stance on religious law: look closely at the actual content of his or her anticorruption platform. For although regime change may be necessary to anticorruption reform, it is clearly not sufficient. Corruption networks are deceptively resilient. Many have survived dramatic efforts to uproot them, ranging from the imprisonment of their leaders to violent revolts against their power. Sometimes they have countenanced their own decapitation, sacrificing a Hosni Mubarak (in Egypt) or a Zine el-Abidine Ben Ali (in Tunisia) to the mob in order to rebound better. In other cases, such as China, they have tried to stay a step ahead of the public by initiating high-profile but self-serving anticorruption measures. By focusing heavily on personal leadership, Rotberg implicitly acknowledges such phenomena without directly grappling with their implications. He thus circles around a crucial question: What is the relationship between kleptocracy and democratic practice? Modern democracy, after all, was developed as a means of guaranteeing government in the public interest. And yet an uncomfortable number of the leaders of Rotberg's "most improved" countries are authoritarians. If firm leadership from the top is so critical to reform, is it even possible for a democracy that has grown systemically corrupt to change course? The United States has become a testing ground for that question. The country's slide into a kind of genteel kleptocracy began many years ago, arguably in the 1980s, when deregulation fever hit. The lobbying profession exploded, and industries began writing legislation affecting their sectors; public services such as incarceration and war fighting were privatized; the brakes on money in politics were released; and presidents began filling top regulatory positions with bankers. An economy of transactional exchanges took hold in Washington. Last year was a watershed in this process. In June, the Supreme Court dramatically narrowed the legal definition of bribery when it overturned the corruption conviction of former Virginia Governor Bob McDonnell. Meanwhile, supporters of the Democratic presidential candidate Hillary Clinton-including many progressive advocates of campaign finance reform-could be heard defending the propriety of questionable foreign donations to the Clinton Foundation. Although Trump's supporters may think otherwise, his victory and ascent to the White House did not represent regime change; they represented very much more of the same, with a president who has invited top corporate executives not merely to provide advice or draft legislation but also to actually join his team. Such a presidency will only cement the system rigging Trump once decried. For Americans, as for the people of so many of the countries Rotberg discusses, the expulsion of one individual at the top will not be enough to repair the damaged republic. Americans should not fool themselves into thinking that all they must do is see Trump impeached or get out the vote for a standard Democratic or Republican alternative in 2020. The network that Trump is anchoring in Washington is exploiting a system that Americans have all allowed to evolve and from which they have averted their eyes. That network is empowered now and will prove resilient.

Endemic delay undercuts efforts to control corruption, for it robs the criminal law of its deterrent effect. The one consistent finding across recent reviews of anti-corruption policy is that it is fear of punishment that dissuades public servants from taking a bribe or committing other corrupt acts (Hanna et al. 2011). If corrupt actors think they can evade consequences by delaying their case, they will discount the threat of punishment. Where court delays are severe, it thus makes little difference how stringent the penalties on the books may be.

Artificial intelligence (AI) is a rapidly growing field of technology with potentially significant implications for national security. As such, the U.S. Department of Defense (DOD) and other nations are developing AI applications for a range of military functions. AI research is underway in the fields of intelligence collection and analysis, logistics, cyber operations, information operations, command and control, and in a variety of semiautonomous and autonomous vehicles. Already, AI has been incorporated into military operations in Iraq and Syria. Congressional action has the potential to shape the technology’s development further, with budgetary and legislative decisions influencing the growth of military applications as well as the pace of their adoption.

There was nothing the doctors could do. In almost total darkness, broken only by the beam of a couple of torches and the glow from their mobile phones, the hospital staff watched helplessly as their patient died in front of them. The elderly woman was suffering a blood clot in her lungs – a common, but life-threatening problem that can be treated with the right drugs and equipment. Everything the doctors needed to save the woman – including a mechanical ventilator – was tantalisingly close, in the intensive care unit several floors below. But with no power in the nine-floor hospital in Maracay, they had no way to reach it. Without electricity the lifts did not work. It was a situation being played out in hospitals dotted all over Venezuela in March 2019 during a that accompanied the growing political and economic crisis facing the South American country. Unprepared for the sudden loss of power, while others only had enough energy to keep a few of the most vital wards functioning. By the end of the five days an estimated 26 people had died in the country’s hospitals as a result of the power outage, according to figures collated by Doctors for Health, a group of concerned medics that have been monitoring the growing health crisis in Venezuela. Among those who died were kidney failure patients who could not get the vital dialysis treatment they needed, and gunshot victims on whom surgeons could not operate in the near darkness. Alongside the deaths were stories of pregnant women giving birth in dark hospital wards, doctors treating patients and surgeons performing operations , and babies in failing incubators. “These babies need special care and without electricity for the incubators staff in neonatal units had to find blankets to keep the babies warm,” says Julio Castro, from the school of medicine at the Central University of Venezuela, who has been compiling the data for Doctors for Health, describing some of the stories that hospital staff had told him about the power outages. “When the ventilators failed, by squeezing a rubber lung,” he says. “They were taking it in turns to keep these patients alive.” The problems extended beyond the hospitals. Elderly people in high-rise flats had to be carried down stairs. People cooked food with fire and ate by candlelight. Without power, food spoiled in warming refrigerators, traffic lights failed and transport systems ground to a halt. The pumps that drove running water to people’s homes stopped, sending residents on a desperate search for . Throughout the year, Venezuela . Some are short and localised, lasting just a few minutes, others take hours for the power to come back, but some go on for days. As they have continued, Castro and his colleagues have recorded more deaths as a result. “If you have even four hours without electricity in a hospital, it is far from normal,” says Castro. “The situation with the water is even worse. There are some hospitals that are having to ask patients to bring in their own water with them because they simply cannot get enough supply.” The situation he describes seems almost apocalyptic in a country that until a few years ago was one of the richest in South America and . Although the Venezuelan government , others have pointed to years of as the cause. But such widespread and long lasting power cuts, known as , are not restricted to countries teetering on the brink of collapse. Each year millions of people in the US and Canada are plunged into darkness by . In June 2019, almost all of Argentina, Uruguay and Paraguay were hit by a power outage that left nearly 40 million people without electricity. In August, almost , trapping commuters on busy trains, when lightning strikes caused a gas-fired power plant and an offshore wind farm to shut down simultaneously. These events, however, are minor in comparison to the kind of power outages that experts fear could be in store in the future. Growing demand on our electricity supplies from rising populations and new technologies like electric cars will face like wind and solar power. Extreme weather events driven by climate change will only heighten the risk to our power supplies further. “So much of our lives and almost everything we do is now dependent on energy, and particularly on our electricity supplies,” says Juliet Mian, technical director of the , an initiative to help organisations and individuals prepare for failures in critical infrastructure. “We used to use the phrase ‘when the lights go out’, but the lights not working are the least of our worries now.” She is right. While the term “black sky” events illustrates perhaps the most visible impact of widespread power failures, it fails to convey the scale of the impact these can have. In our modern world, almost everything, from our financial systems to our communication networks, are utterly reliant upon electricity. Other critical infrastructure like water supplies and our sewer systems rely upon electric powered pumps to keep them running. With no power, fuel pumps at petrol stations stop working, road signs, traffic lights and train systems go dead. Transport networks grind to a halt. Our complex food supply chains quickly fall apart without computers to coordinate where produce needs to be, or the fuel to transport it or refrigeration to preserve it. Air conditioning, gas boilers and heating systems also rely upon electricity to work. A little over 100 years ago, our cities ran on human and animal muscle power to ferry goods and waste around. Modern infrastructure is now utterly reliant upon electricity. “In today’s world, our systems are highly interdependent and it is very hard to find many systems that are not fundamentally reliant upon power,” says Mian. “A black sky scenario will affect everyone.” The causes of a black sky event are many. They vary from natural disasters like hurricanes or earthquakes to , or coronal mass ejections, that send a barrage of electrically charged particles racing across the Solar System and . One intense geomagnetic disturbance caused . The Electric Infrastructure Security Council, an international body that reviews threats to power grids, also . These include cyberterrorism attacks or coordinated physical assaults on energy infrastructure such as power stations, and electromagnetic pulses that can disable electricity grids. “Our national power grids are tremendous feats of engineering and operations that have supported rapid economic growth around the world,” say Melissa Lott, a research fellow at the Center on Global Energy Policy at Columbia University in New York. “But more investment is needed if electric power grids are going to keep up with rapid technology shifts and increasingly extreme weather events.” She says that while true black sky events are mercifully rare, the deep impact they have on businesses and people means more needs to not only update grid technology and management, but also improve infrastructure so it can be more resilient against physical threats like flooding. “In the summer of 2012, blackouts in India cut power to more than 600 million people over two days. In Puerto Rico, Hurricane Maria crippled infrastructure across the island, leaving people in the dark and triggering a humanitarian crisis. In 2018, an earthquake on Japan’s Hokkaido island left more than 5 million people without power. In order to keep these events from becoming more common and to minimise their impact, we need to invest in our grids.” Putting measures in place to counter all of these potential threats is difficult and expensive. Critical systems can be guarded from human attacks and they can be shielded from electromagnetic pulses with enough money being spent on them. Building new systems for can also help to keep systems safe. But there are some events that cannot be planned for and the complex, interconnected nature of our electricity grids are remarkably vulnerable. Take what happened in September 2003 when a fallen tree brought down a power line in Switzerland’s Lukmanier Pass over the Alps into Italy and 24 minutes later another tree came down onto a line in the nearby Great St Bernard pass. The sudden failure of these two key lines caused other connections to Europe’s electricity network to trip, which triggered . The whole of Italy was left without power because of two fallen trees starting a cascade of events. Modern electricity grids are increasingly interconnected and complicated, making failures like this difficult to predict. Most of Europe now runs off – probably the largest in the world – that supplies more than 400 million customers in 24 countries. The USA is made up of five different grids. But there are some that are seeking ways of anticipating potential power failures and are enlisting the help of artificial intelligence to help them grapple with this highly complex problem. When a power plant goes down, for example, it causes an abrupt spike in load on others on the network, which in turn slows down the generators at these plants and causes the frequency held on the grid to decrease. This risks destabilising the delicate balance that electricity grids are held in, and operators have to deploy countermeasures rapidly – often within milliseconds – to prevent sections of the grid being cut off. Researchers at the Fraunhofer Gesellschaft research institute in Ilmenau, Germany, recently revealed they are developing an AI system to and take steps to address them. The US Department of Energy is also funding $7m (£5.4m) of research into using artificial intelligence to not only that might lead to larger problems, but also to help find ways of keeping supplies constant in the event of a problem. General Electric is using machine learning to help analyse weather forecasts past outage history and information on the ground from its response crews to predict the impact that impending storms might have on its networks. It is also using it to predict where its repair crews might need to be so that downed lines can be restored more quickly. Power grids can also help to protect themselves by they have available so that supplies can be supplemented when generators do go off-line unexpectedly. But completely protecting our power networks from failures is almost impossible, says Mian. “We can’t necessarily design our systems so that failures don’t happen,” she says. “There’s so much complexity in our systems these failures can cascade and they can become much more widespread, which means the failure is quite often unavoidable. But what we can do is design our systems so that they can respond and recover quickly.” This is what the Resilience Shift is now trying to improve. It has been organising exercises in collaboration with the Electric Infrastructure Security Council (EIS) that help large organisations, universities, schools, community groups and even families prepare themselves for an event that might lead to mass power cuts for several days at a time. The Emergency All-sector Response Transnational Hazard Exercise, or Earth Ex, is an online exercise that allows people to rehearse the decisions they need to make and put in place the plans they need should the worst happen. () “We want people to be thinking about these things long before there is a problem,” says John Heltzel, director of resilience planning at the EIS. “It’s important because when the electricity grid fails, there is this risk of cascading impacts that can occur from what might at first seem to be a relatively minor event. This cascade effect is where the real damage can be done. As the people of Venezuela have discovered, even basic service like water can stop when the power goes out. “It’s effectively [like going] back to the dark ages,” says Heltzel. A report by scientists at University College London mapped out how the , from the loss of health care provision and sanitation services to citizens trapped in lifts and disruption to transport systems. Then there are the social consequences that percolate out. as they provide opportunities for theft and fraud. The supply of cash and credit – particularly in our modern societies so reliant upon electronic and card payments – dry up meaning people have to rely on whatever cash they happen to have squirreled away. Communication networks and the ability to contact loved ones disappear, while vulnerable people like the elderly are often left stranded in their own homes. Businesses are also left largely unable to operate, resulting in huge economic impacts. In 2004, the Department of Energy estimated the annual cost of power outages in the US . When two million customers in California had their supplies cut for two days in October this year, experts estimated the cost to the economy to be around $2.5bn (£1.9bn). Heltzel knows first hand the kind of chaos large-scale power outages can cause. He is a retired brigadier general who spent 33 years in the Kentucky National Guard and also served as the deputy commander of the Kentucky Joint Force Headquarters. In 2009, the state was hit by a series of ice storms that brought powerlines tumbling down under the weight of rime ice and snow building up on the wires. “On one day we got an ice storm, followed by a snow storm, followed by another ice storm,” recalls Heltzel. The build up of ice was so great it even brought down steel utility structures designed to withstand hurricane force winds and , according to a later Congressional hearing. “We lost power in all of western Kentucky,” says Heltzel. “From a state perspective, of 120 counties, we had 114 placed into a state of emergency. It meant that people were stuck in their houses and couldn't get to the stores to buy food. So, we had people that were going hungry and we had people whose wells were frozen over. They could not get water through the normal municipal water systems. At the same time the communications networks weren't working, so they could not call for help.” The Kentucky National Guard mobilised 12,000 soldiers and airmen to go door-to-door delivering food to people. They also requested emergency generators to get the water supply back up and running. Emergency communication stations were brought in from other states to restore the telephone and radio network. Even so, the hardest hit areas were . “We were flying people who managed the electricity grid up and down their power lines in our helicopters so they could assess how many poles and cables they needed,” says Heltzel. “But even with all the resources we brought to bear, it took four and half weeks for the last house to be reconnected.” Around 35 people in Kentucky and 30 in neighbouring states lost their lives. At least eight of the deaths were due to without proper ventilation. This is why Heltzel believes planning for a black sky event before one happens is so important. Organisations like hospitals, water suppliers and large companies can make sure they regularly service their back-up generators, and have a sufficient supply of fuel to keep them running. Churches and schools can make sure they have blankets on hand and other facilities to help those who might become stranded and need shelter. On an individual level, we can all take steps too. From simple things like having torches with plenty of spare batteries to hand, to ensuring we have adequate supplies of bottled water to fall back on – the EIS recommends having two weeks worth of water with two litres a day per person and one litre for pets. Keeping cupboards stocked with non-perishable foods like rice, pasta and canned vegetables, is also advised. But the Heltzel and his team also have some more unusual tips for families looking to prepare for a black sky event. Baby formula, for example, is a good source of nutrition even if you don’t have young children. A good supply of rubbish bags is also important – these can be tucked under the toilet seat if the water stops running, allowing you to bury your waste outside. Keeping a stash of emergency cash could also be a life saver. “One of the things we talk about for individuals and families is turning yourself from being a survivor into someone who can help with the restoration,” adds Heltzel. “We want people to be part of the solution rather than the problem. That might be being part of wider community efforts to build resilience or simply helping others who are not as prepared.” In Venezuela, the medical staff have become a perfect example of this. Following the first nationwide black outs, the number of deaths in hospitals have declined with each subsequent black out. Julio Castro, from Doctors for Health, puts this partly down to the shorter length of the power outages, but also to the preparations that hospital staff put in place. “Now they are aware of the problem they have put procedures in place,” says Castro. “They have made sure they have fuel and the back up generators are working. They have rotas for when they need to do manual ventilation and manual back ups for their equipment. “It is keeping people alive.”

Consequences of a sustained power outage The EMP Commission states “Should significant parts of the electrical power infrastructure be lost for any substantial period of time, the Commission believes that the consequences are likely to be catastrophic, and many people will die for the lack of the basic elements necessary to sustain life in dense urban and suburban communities.” [67]. Space constraints preclude discussion on how the loss of the grid would render synthesis and distribution of oil and gas inoperative. Telecommunications would collapse, as would finance and banking. Virtually all technology, infrastructure, and services require electricity. An EMP attack that collapses the electric power grid will collapse the water infrastructure—the delivery and purification of water and the removal and treatment of wastewater and sewage. Outbreaks that would result from the failure of these systems include cholera. It is problematic if fuel will be available to boil water. Lack of water will cause death in 3 to 4 days [68]. Food production would also collapse. Crops and livestock require water delivered by electronically powered pumps. Tractors, harvesters, and other farm equipment run on petroleum products supplied by an infrastructure (pumps, pipelines) that require electricity. The plants that make fertilizer, insecticides, and feed also require electricity. Gas pumps that fuel the trucks that distribute food require electricity. Food processing requires electricity. In 1900, nearly 40% of the population lived on farms. That percentage is now less than 2% [69]. It is through technology that 2% of the population can feed the other 98% [68]. The acreage under cultivation today is only 6% more than in 1900, yet productivity has increased 50 fold [69]. As stated by Dr. Lowell L Wood in Congressional testimony: “If we were no longer able to fuel our agricultural machine in the country, the food production of the country would simply stop, because we do not have the horses and mules that used to tow agricultural gear around in the 1880s and 1890s”. “So the situation would be exceedingly adverse if both electricity and the fuel that electricity moves around the country……… stayed away for a substantial period of time, we would miss the harvest, and we would starve the following winter” [70]. People can live for 1–2 months without food, but after 5 days, they have difficulty thinking and at 2 weeks they are incapacitated [68]. There is typically a 30-day perishable food supply at regional warehouses but most would be destroyed with the loss of refrigeration [69]. The EMP Commission has suggested food be stockpiled for a possible EMP event. A prescription for failure Even if all the recommendations of the Congressional EMP Commission were implemented, there is no guarantee that the grid will not sustain a prolonged collapse. There should therefore be contingency plans for such a failure. There is also another consideration. The foundational pillars of prior American nuclear defense policy, in today’s climate, are of uncertain validity. Mutual assured destruction is the Maginot line of the 21st century. Nonproliferation will prove difficult to resurrect. The consequences of a widespread nuclear attack have been positioned to the public as massive deaths from blast effects, and then further lingering deaths from the effects of radiation. We suspect there will be no electricity, and there will be no electricity for a very long time. There should be an actionable plan in anticipation of a possible prolonged collapse of the grid—a retro-structure and a skill set to provide a framework for survival. Our sense is there

It has been shown previously that the four integral components of the Earth's climate system mainly considered here are at risk of transgressing into undesirable states when critical thresholds are crossed (Schellnhuber et al., 2016; Lenton et al., 2008). Over the past decades, significant changes have been observed for the polar ice sheets as well as for the Atlantic Meridional Overturning Circulation (AMOC) and the Amazon rainforest (Lenton et al., 2019). Should these climate tipping elements eventually cross their respective critical temperature thresholds, this may affect the stability of the entire climate system (Steffen et al., 2018).

In this study, we show that this risk increases significantly when considering interactions between these climate tipping elements and that these interactions tend to have an overall destabilising effect. Altogether, with the exception of the Greenland Ice Sheet, interactions effectively push the critical threshold temperatures to lower warming levels, thereby reducing the overall stability of the climate system. The domino-like interactions also foster cascading, non-linear responses. Under these circumstances, our model indicates that cascades are predominantly initiated by the polar ice sheets and mediated by the AMOC. Therefore, our results also imply that the negative feedback loop connecting the Greenland Ice Sheet and the AMOC might not be able to stabilise the climate system as a whole, a possibility that was raised in earlier work using a Boolean modelling approach (Gaucherel and Moron, 2017).

Several epidemics, such as the Black Death and the Spanish flu, have threatened human life throughout history; however, it is unclear if humans will remain safe from the sudden and fast spread of epidemic diseases. Moreover, the transmission characteristics of epidemics remain undiscovered. In this study, we present the results of an epidemic simulation experiment revealing the relationship between epidemic parameters and pandemic risk. To analyze the time-dependent risk and impact of epidemics, we considered two parameters for infectious diseases: the recovery time from infection and the transmission rate of the disease. Based on the epidemic simulation, we identified two important aspects of human safety with regard to the threat of a pandemic. First, humans should be safe if the fatality rate is below 100%. Second, even when the fatality rate is 100%, humans would be safe if the average degree of human social networks is below a threshold value. Nevertheless, certain diseases can potentially infect all nodes in the human social networks, and these diseases cause a pandemic when the average degree is larger than the threshold value. These results indicated that certain infectious diseases lead to human extinction and can be prevented by minimizing human contact. 1. Introduction The emergence of a pandemic is one of the various scenarios frequently discussed as a human extinction event, and it is listed as one of the global catastrophic risks in studies regarding the future [1,2,3]. In particular, several pandemics, such as the Black Death [4,5], Spanish flu [6], and those caused by smallpox [7], severe acute respiratory syndrome (SARS) [8], and Ebola [9], have affected a large population throughout history. The risk of pandemics increases with an increase in population mobility between cities, nations, and continents, thereby threatening humankind [10,11,12]. It is essential to analyze the epidemic spread in society to minimize the damage from epidemic disasters; however, extinctive epidemic spreading experiments have limitations in real-world situations, as they predict stochastic effects on the spread without considering the structure of human society. Network-based approaches have been proposed to overcome these limitations and perform epidemic spreading simulations by considering the network structure of numerous real-world connections [13,14,15]. These methods use various models of epidemic spreading, such as the susceptible–infectious–susceptible (SIS) [16,17,18], susceptible–infectious–recovered (SIR) [19,20,21], and Watts threshold models [22]. While these methods are mathematically convenient, they are epidemiologically unrealistic for various infections because they require exponentially distributed incubation and infectious periods [23,24,25]. Moreover, previous epidemic studies did not perform quantitative assessment of the pandemic risk depending on the network connectivity in individuals and fatality rate of various diseases [26]. In the present study, we applied an SIR epidemic model to a scale-free network with Monte Carlo simulation to identify the quantitative relationship between infectious diseases and human existence. Our fundamental hypothesis states that when the epidemic spreads to all nodes of the network and the fatality rate is 100%, it can increase the pandemic risk. To address this, we initially constructed a scale-free network to simulate a society. Moreover, for the epidemic spreading simulation, an SIR model was applied to the network to describe the immune state of an individual after infection. From the simulation study, we found that the mean degree of a scale-free network was an essential factor in determining whether epidemics threaten humans. This approach provides important insights into epidemic spreading analysis by investigating the relationship between epidemic and scale-free network parameters. Furthermore, it highlights the necessity of determining information flow during an epidemic. 2. Materials and Methods We designed an epidemic simulation process to identify the relationship between pandemic risk and network parameters. This study was performed in four steps (Figure 1): (i) generating a scale-free network model to reflect real-world conditions; (ii) applying an SIR model to the scale-free network for epidemic spreading simulations; (iii) adapting the Monte Carlo method to reflect the stochastic process in the node status of the SIR model; and (iv) iteratively performing simulation for every parameter set and analyzing the results. We have provided the source code and sample results of epidemic simulation in Supplementary Materials.

Figure 1. Overview of epidemic simulation process based on the Monte Carlo method. (A) We generated scale-free networks for a fixed population (N = 1,000,000) and various node degrees (k = 2, 5, 7, and 10). (B) Epidemic spreading was simulated by applying a susceptible–infectious–recovered (SIR) model to the scale-free network. We set the epidemic parameters, β and γd. β represents the spreading rate of epidemics, and γd is the reciprocal of γ and reflects the time interval between infection and recovery. Randomly, 0.05% of nodes were initially infected. (C) We adapted the Monte Carlo method to determine the status of the transition from the infection node to immunization node. Repeated simulations were performed until a steady state was achieved. (D) For every parameter set, 10,000 simulations were performed. 2.1. Network Generation Based on a Scale-Free Model We constructed a network model for the epidemic spreading simulation (Figure 1). The nodes and edges of the network represent people in the society and their physical contacts, respectively. We used a scale-free network model, which follows the preferential attachment property observed in numerous real-world networks, such as social networks, physical systems, and economic networks [27,28,29]. In the scale-free network, when a node is added to the network, its likelihood of connecting to existing nodes increases with an increase in the node’s degree. Hub nodes, which lead to fast and vast spreading of epidemics, exist. Two characteristic parameters, including N and k, affect the form of scale-free networks. The parameter N denotes all nodes in the network. In the real world, N indicates the whole population size. The parameter k is the average degree of the network, which determines the degree of the newly attached node for each step during network generation. Following the characteristics of the network model, we generated scale-free networks representing human contacts for epidemic spread. The scale-free network was generated by the Barabasi–Albert graph distribution, in which the network is constructed from a cycle graph with three vertices, followed by the addition of k edges at each construction step [30]. The k edges are randomly attached to the vertex based on the degree distribution of the vertex. After network generation, we investigated the degree distribution properties of the network (Figure 2). The results indicate that the degree distributions have similar tendency for networks with varying number of nodes and edges. This study constructed scale-free networks with the largest number of nodes considering computational complexity (N = 1,000,000). [FIGURE 2 OMITTED] Figure 2. Degree distribution of the scale-free network. We analyzed the degree distribution of the network based on the number of nodes (N) and mean degree (k). 2.2. Epidemic Spreading Based on the SIR Model For the epidemic spreading simulations, we applied an SIR model to the generated scale-free network. The classical SIR model can be expressed by the following nonlinear differential equations [21]: [EQUATIONS OMITTED] where S, I, and R represent susceptible, infected, and recovered compartments, respectively, in the whole population. S represents people who have not been infected yet but can be infected in future. I represents infected people who can spread the epidemic to susceptible people through physical contact. R denotes people who have recovered or died from the epidemic and who no longer participate in the epidemic spreading process. The sum of the S, I, and R values represents the whole population size N. Epidemics have two parameters in the SIR model, transmission rate (β) and recovery rate (γ), which arise from the basic reproduction number R0 (Figure 1B). The basic reproduction number is the number of infections caused by one infective node [31,32,33]. If the R0 is more than 1, the infection can spread in a population, whereas if R0 is less than 1, the infection cannot spread. We express the basic reproduction number as R0 = β/γ, where β represents the spreading rate of epidemics between infective nodes and adjacent susceptible nodes and γ represents the probability of recovery from infection [34]. We mainly used γd, which is the reciprocal of γ and reflects the time interval between infection and recovery. 2.3. Investigation of Epidemic Status Based on the Monte Carlo Method The epidemic simulation was performed for a time series event by constructing epidemic status matrix (z) to represent the status of the nth node at time step t. For each node, the value of epidemic status matrix at time step t can be 0, 1, or 2, indicating that a node is susceptible, infective, or recovered, respectively. We initially (t = 0) set every value of epidemic status matrix to 0 because all nodes are susceptible before the epidemic spreads. At the initial infection stage, randomly selected 0.05% of nodes were infected. At every time period, we performed immunization and observed the infection stages (Figure 3). [FIGURE 3 OMITTED] At the immunization stage, we identified infective nodes and determined whether these nodes would be recovered in the next time step. To calculate the transition probability of infected and recovered phenomena, the Monte Carlo method was applied [35,36]. When infection and recovery parameters are provided, it is possible to investigate whether a node transitions from an epidemic state to another state. To accomplish this, we compared the method revealing the change in each population in every compartment over time (Figure 4). [FIGURE 4 OMITTED] The final steady state of the epidemic spreading simulation model indicates the total number of casualties of the epidemic who either are dead or have recovered from the disease. Infective nodes at time t (zn [t] = 1) are transformed to recovered nodes at time t + 1 (zn [t + 1] = 2) when 1/γd is larger than a random real number between 0 and 1. We determined whether the neighbor nodes of the infection node would be infected by identifying susceptible nodes adjacent to the infective nodes at time t (zn [t] = 0, with the adjacent infective node) (Figure 5). When β is larger than a random real number between 0 and 1, a susceptible node becomes an infective node at time t + 1 (zn [t + 1] = 1); this scenario represents epidemic spread. For each time step, we recorded the number of susceptible, infective, and recovered nodes during epidemic spread. [FIGURE 5 OMITTED] 2.4. Simulation Parameters We carried out simulation trials for various mean degrees of networks (k = 2, 5, 7, and 10). Each network considered the following epidemic parameters: β ranges from 0.05 to 0.95 and γd ranges from 1 to 10. The Monte Carlo model was repeatedly simulated to observe saturation of the recovery process. Considering that the simulation pipeline contains random processes such as initial infection and Monte Carlo trials, we performed the simulation iteratively until the status of nodes remained unchanged. After simulation, time series data from every simulation were interpolated in the time domain. The fatality rate determines the ratio of deceased and recovered individuals in the final population [37,38,39]. If the fatality rate is below 100%, the recovered population contains both dead and recovered individuals. Such a situation does not always cause a pandemic. In this simulation, we assumed a 100% fatality rate. To accomplish this, we enumerated the recovered nodes as dead for considering the pandemic risk. 3. Results Through our method, we obtained epidemic spreading data with various network and epidemic parameter sets. In the present study, we focused on the case where the epidemic infects all nodes and defined this phenomenon as “extinctive spread”. Diseases causing extinctive spread are potential candidates of high pandemic risk. In the real world, extinctive spreading indicates that the disease will infect every person in the society. From the simulation data, we calculated the extinctive spread score by dividing the total number of simulation trials by the number of extinctive spread cases. Thereafter, we identified that the number of extinctive spread cases is mainly influenced by spreading speed, which is determined by β, γd, and k (Figure 6). [FIGURE 6 OMITTED] The extinctive spread region (brown area in Figure 6) is expanded as the value of mean degree of network (k) is increased, thereby indicating that the area of extinctive spread becomes noticeably wider in a dense network than in a sparse network. Thus, the more contact between people, the higher the risk of epidemics. Moreover, high γd and high β cause extinctive spread across a large region, indicating that the high spreading rate and short time interval between infection and recovery are risk factors of epidemic diseases. In contrast, the infective nodes recover before they transmit the disease to their neighbors in low β and low γd scenarios, thus disconnecting the network and preventing extinctive spread. This occurs because the infective nodes need more time to transmit the disease in low β and high γd scenarios. Therefore, the disease begins to subside due to a lack of new infective nodes. Furthermore, we investigated the range of β and γd for existing epidemics of the common cold [40,41] and fatal diseases, namely, cholera [42,43], Marburg [44,45], Ebola (Congo and Uganda) [46,47,48,49], SARS [50], and MERS [51] (Table 1). We selected diseases with relatively well-known epidemic parameters, such as average duration of infection and basic number of reproductions from previous studies. Transmission rates were calculated using the mean duration of infectious periods and basic reproduction numbers of the epidemics. Different studies reveal multiple values of infectious period and transmission rate for some of these diseases; we considered these values separately [40,41,42,43,46,47,48,49]. For example, the infectious period of a common cold is from 3 to 7 days and that of Ebola is 6.5 days. Next, we placed the possible regions of these epidemics as a disease band for various k values (colored lines in Figure 6). When k > 5, fatal diseases have an opportunity to cause a pandemic. Even when k = 5, diseases such as cholera and Ebola (Congo) can be threatening in regions of low γd and high, thus demonstrating that the knowledge of network parameters of the society and the characteristics of epidemic diseases can aid in quantifying the risk of epidemics. [TABLE 1 OMITTED] 4. Discussion Many previous studies have made stochastic SIR models to analyze the dynamics or stability of epidemic diseases. They investigated the distribution of susceptible, infected, and removed populations for specific epidemic disease spreading, such as cholera, SARS, Marburg, and MERS, based on mathematical modelling [52,53,54,55]. However, they did not conduct a quantitative assessment of pandemic risk taking into account physical contact between people. To solve this limitation, we performed epidemic spreading simulations by applying an SIR model to scale-free networks with Monte Carlo simulation. In the simulation, we consider various connectivity and disease characteristics on scale-free networks. For each network and epidemic parameter set, the probability of extinctive spread was calculated. The results revealed that certain infectious diseases can lead to extinction. Moreover, even if the disease band extends over the extinctive spread regions, it does not indicate that human extinction results from the disease, as the fatality rate is below 100%; however, in the case of 100% fatality, the disease can cause a human extinction event. The risk of infectious disease is influenced by the network structure. A dense network has a higher risk of spreading infectious disease than a sparse network, as we observed in the extinctive spreading maps. According to our results, when the average degree of human social networks is below the risk threshold, i.e., less than 4 in this study, human society is safe from an extinctive outbreak based on our knowledge regarding the epidemic parameters of the infectious disease. Nevertheless, in other cases, human extinction is possible. For example, if the population is 1,000,000 and there are 4 or more instances of physical contact between people, human extinction events may occur, depending on the fatality rate of the epidemics. Hence, physical contact between people is closely related to an extinction event of infectious diseases. Eventually, from a public health perspective, lowering the average contact level of society is an appropriate way to increase the robustness of strategies against the occurrence of extinction. In the real world, reducing network density can be accomplished by epidemic prevention activity, such as isolation and quarantine treatment. This action prevents epidemic risk to the society, thereby avoiding human extinction. Additional considerations may improve our analysis. First, large population size and various proportions of initial infective nodes were not considered in the experiments. We have confirmed that the result was consistent when the proportion of initial infective nodes was 0.05% of the total population; however, this can vary depending on the distinct proportion of initial infective nodes in a different population. To achieve robust results, we need to perform additional experiments for various parameters; however, we could not address this issue due to computational complexity. Second, we did not consider numerous known epidemic diseases. We calculated the transmission rates of epidemic diseases using the known infectious periods and reproduction numbers of the epidemics from evidence in the literature. In the present study, we only considered five epidemic diseases, since the information on infectious periods and reproduction numbers of diseases was mostly unavailable for other epidemic diseases. Third, this study only considers the SIR model on scale-free networks in epidemic simulation. Since the dynamics of epidemic diseases can be varied in different models or networks, it is important to experiment in various simulation environments to confirm the robustness of the results. Nevertheless, these limitations can be considered in future experiments or using improved computational methods. With these further improvements, our approach can be used as a computational tool to analyze the risk of epidemic diseases. 5. Conclusions In this study, we analyzed the risk of epidemic diseases by creating an epidemic simulation on a scale-free network. Based on the simulation results for various epidemic parameters, we confirmed that certain infectious diseases can lead to extinction and can be prevented by minimizing human contact. We believe that identifying potential candidate diseases that may lead to human extinction is crucial in addressing epidemic prevention activities such as quarantine.

2 Grim situation and consensus on human activities changing climate A team consist of 93 scientists has published a exceptionally comprehensive record of paleoclimatic data across the past 12,000 years. It contains 1,319 data records collected from 679 sites around the world and from samples such as marine sediments, lake sediments, peat, coral, cave sediments and glacial ice cores. From this data, the researchers mapped changes in surface air temperatures over the 12,000 years since the last iceage. The figure is compared to the century average temperature between 1800 and 1900 to track changes likely to be brought about by the Industrial Revolution. As expected, temperatures at the start of the period were much cooler than the 19th century baseline. But over the next few thousand years, temperatures rose steadily, eventually surpassing the baseline. Temperatures peaked 6,500 years ago, and since then, the planet has been slowly but surely cooling, seemingly driven by slow cycles in Earth's orbit, which reduced the amount of sunlight in the northern hemisphere's summer and ultimately led to the 'Little Ice Age' of recent centuries. In a relatively short period since the middle of the 19th century, human activity has increased the average temperature by as much as 1℃, a huge peak that is higher than the peak of 6500 years ago (Fig.1) [1].

[figure excluded]

It is tempting to think of climate change as a gradual, linear process that follows the steady accumulation of greenhouse gasses in the atmosphere. The truth may be scarier. Evidence is mounting that critical components of the Earth system could be approaching tipping points that, when crossed, will bring about massive, nonlinear shifts that will themselves accelerate climate change, with disastrous and potentially irreversible consequences for nature and humanity. Faced with this prospect, national governments must cooperate to identify, prepare for and, where possible, head off or mitigate these abrupt and calamitous shifts.

As the Intergovernmental Panel on Climate Change made clear in its recent assessment report, the long-foreseen climate emergency is upon us. For the past 12,000 years, a period geologists call the Holocene, humanity has enjoyed remarkable climate stability. That benign era, which permitted the emergence of civilization, is over. The global environment is changing in ways unprecedented in human history. Atmospheric carbon dioxide levels are the highest they’ve been in 800,000 years, and the past decade has included nine of the 10 hottest years on record. Glaciers and arctic sea ice are retreating; the oceans are warming and acidifying; climatic zones are shifting poleward; sea levels are rising; species and ecosystems are disappearing; and searing heat waves, punishing droughts, heavy precipitation and tropical cyclones are increasing in frequency and intensity.

Geography and distance are key factors in determining plans and strategy, as well as broader security and diplomatic policies of states (Heginbotham et al. 2015). Furthermore, the BoT theory suggests that rising powers like China, geographical proximity “will inevitably induce fear and apprehension among surrounding smaller states” (Chu et al. 2015, p. 406).

Nonetheless, the concept becomes obsolete in the face of the ETs that transcend borders, with threat perceptions no longer confined to surrounding or neighboring states. China’s pursuit of global digital leadership in multiple areas, such as digital infrastructure, e-commerce, and research collaboration (Shi-Kupfer and Ohlberg 2019), has been evident in the development of a global Chinese technology network through state-backed companies. Chinese companies supply AI surveillance technology to 63 countries that use exclusively Chinese technology, most of which are in Africa and Latin America (Feldsten 2019). At the same time, state-backed companies such as Huawei, Alibaba, and Tencent have already established their presence in numerous states' telecommunications networks, data centers, and online payment systems. Through initiatives like the Digital Silk Road (DSR), these companies, including Huawei, ZTE, Hikvision, NucTech, and others, are exporting AI technologies, providing support, and directly participating in the deployment and development of various technologies such as cloud computing, smart cities, digital payment systems, and AI within host countries (Greene and Triolo 2020; Feldstein 2019). As this intricate web of digital market interconnections continues to expand, it presents a substantial challenge to other powers irrespective of their geographical locations (Shi-Kupfer and Ohlberg 2019). As such, Beijing’s investments in several countries allow China to leverage its global technology power image (Fannin 2020).

The US government has repeatedly raised its concerns regarding potential security risks from these companies due to the influence of the Chinese government over its firms, proclaiming that the Chinese government “may force Chinese suppliers or manufacturers to modify products to perform below expectations or fail, facilitate state or corporate espionage, or otherwise compromise the confidentiality, integrity, or availability of IoT devices” (USCC, 2018, p.461). Given China’s notorious cyberattacks reputation, new tensions pertaining to the AI-cyber nexus may emerge in the near future. Cyber tactics such as cyber espionage and attacks have become a cost-effective modus operandi for states, especially China, in relation to ETs and offer a means to acquire otherwise inaccessible or heavily guarded information, providing a substantial advantage in the rapid advancement of these technologies. According to US think tank, the Center for Strategic and International Studies, between 2000 and 2023, 224 incidents were recorded, with 49% involving actors tied to the Chinese military and government. Moreover, over 1200 cyber espionage cases in either China or the US are currently under investigation (CSIS 2023a, b). Some estimates propose that the US economy suffers an annual loss of US$300 billion due to Chinese cyber espionage operations, affecting areas from quantum computing and semiconductors to AI (Laskai and Segal 2018). Over time, there have been long-standing allegations of cyber espionage efforts by China along with forced technology transfers through mergers with Western companies, which have raised significant concerns among the US and its allies and partners (Lindsay et al. 2015). At the same time, in its 2022 Annual Threat Assessment, the Office of the Director of National Intelligence (ODNI) included the following passages regarding China’s cyber capabilities and intentions: “China presents the broadest, most active, and persistent cyber espionage threat to US Government and private sector networks. China’s cyber pursuits and export of related technologies increase the threats of attacks against the US homeland. China almost certainly is capable of launching cyberattacks that would disrupt critical infrastructure services within the United States, including against oil and gas pipelines and rail systems” (ODNI 2022, 8). Between 2005 and 2022, 237 cyber operations carried out by China (state sponsored) were reported (according to Council of Foreign Affairs Cyber Operations Tracker 2023), which can be interpreted as aggressive actions of malicious intent.

Similarly, China’s disinformation campaigns reinforce the assessment of China’s intentions as aggressive. Since 2015, has expanded social media disinformation campaigns, and established a separate unit for the conduct of information warfare, the PLA Strategic Support Force (PLASSF) (Harold et al. 2021). While China has been mainly using information manipulation domestically, for instance, during the Hong Kong protests using Western social media such as Twitter (Beskow and Carley 2020), its latest extensive disinformation campaigns have posed a significant risk to US and other actors’ intelligence operations and national security. Indeed, China (similar to Russia) has been accused of spreading disinformation campaigns and propaganda amid the COVID-19 crisis by both the EU and the US (European Parliament 2020; US Embassy in Georgia 2020), with the goal of advancing “anti-Western views, and spread[ing] false information to create divisions between the United States and its partners and allies” (White House 2020, p. 7).

This element is particularly important to emerging powers since the offensive capabilities can be used in arms forces. The use of AI, robotics, directed energy, cyber warfare, and orbital systems secure competitive advantages and strengthen deterrence by curbing lower level challenges and achieving strategic advantage. China has developed several systems that could penetrate US defenses. According to some accounts, "China has likely surpassed the US in the development of a new class of weapons, hypersonic strike vehicles, and may be within months of deploying such systems" (Lewis 2019, p. 6). In addition, the focus is not on military expenditure or the number of military equipment but on the ability to significantly disrupt, degrade, and dismantle military infrastructure.

One could reasonably argue that China’s development and enhancement of offensive capabilities position it as a potential threat, even though its defense budget does not align with that of the US.Footnote22 Nevertheless, Xi Jinping has pushed, as indicated in the “Military Strategic Directive for a New Era”,Footnote23 for intensifying innovation in ETs to “build a strong military” and reach greater strategic autonomy, while the 14th 5-year plan (2021–25)Footnote24 describes a goal of developing disruptive technologies that would give China a military advantage over rivals and close the gap with the US (Duchâtel 2023).

According to the Center for Strategic and International Studies, China is a near-peer competitor with the US in AI and advanced computing (Lewis 2019). Some authors even consider China to be ahead of the US and at the forefront, globally, in some fields such as quantum computing (Whalen 2019). Following the idea of threat perceptions intertwined with a change in material capabilities, and examining China’s “aggregate power” in ETs, we can observe that the country has made significant advances in various areas while still lagging behind the US in overall technological and research performance (Paszak 2019; Brooks and Wohlforth 2016). This is particularly true of semiconductors (Levine 2020).

Nevertheless, the country’s growing and expansive relative ET capabilities, advancements, and incremental growth over a short period of time have positioned it as a threat toward the US. This is supported by US officials’ accounts of the ETs, who have stated that “in many ways, it is less important whether their technology is ‘as good as ours’ than whether it is good enough to render our capabilities ineffective” (Carter 2018, p. 5). Former US Secretary of Defense Mark Esper also voiced the widely shared concern: “I think we need to be very concerned about Chinese technology getting into our systems or the systems of our allies” (Macias 2019) (Table 2).

Despite 30 y of efforts and some progress under the United Nations Framework Convention on Climate Change (UNFCCC) anthropogenic greenhouse gas (GHG) emissions continue to increase. Even without considering worst-case climate responses, the current trajectory puts the world on track for a temperature rise between 2.1 °C and 3.9 °C by 2100 (11). If all 2030 nationally determined contributions are fully implemented, warming of 2.4 °C (1.9 °C to 3.0 °C) is expected by 2100. Meeting all long-term pledges and targets could reduce this to 2.1 °C (1.7 °C to 2.6 °C) (12). Even these optimistic assumptions lead to dangerous Earth system trajectories. Temperatures of more than 2 °C above preindustrial values have not been sustained on Earth’s surface since before the Pleistocene Epoch (or more than 2.6 million years ago) (13).

Even if anthropogenic GHG emissions start to decline soon, this does not rule out high future GHG concentrations or extreme climate change, particularly beyond 2100. There are feedbacks in the carbon cycle and potential tipping points that could generate high GHG concentrations (14) that are often missing from models. Examples include Arctic permafrost thawing that releases methane and CO2 (15), carbon loss due to intense droughts and fires in the Amazon (16), and the apparent slowing of dampening feedbacks such as natural carbon sink capacity (17, 18). These are likely to not be proportional to warming, as is sometimes assumed. Instead, abrupt and/or irreversible changes may be triggered at a temperature threshold. Such changes are evident in Earth’s geological record, and their impacts cascaded across the coupled climate–ecological–social system (19). Particularly worrying is a “tipping cascade” in which multiple tipping elements interact in such a way that tipping one threshold increases the likelihood of tipping another (20). Temperature rise is crucially dependent on the overall dynamics of the Earth system, not just the anthropogenic emissions trajectory.

For years, the two of us have been warning of the dangers to America’s startup system posed by big startup losses. In 2013, angel investor Aileen Lee coined the term “unicorn” for startups that were valued at $1 billion or more before they went public. We have been documenting the unprofitability of these unicorns along with their huge cumulative losses (see here, here, and here). Furthermore, almost 60% of America’s publicly traded unicorns have cumulative losses larger than their revenues — a figure that doesn’t include the dozens of publicly traded unicorns which have gone bankrupt or been acquired in the past two years. Privately traded startups are almost surely doing even worse, because the best will IPO first. Thus, Morgan Stanley’s European head of research reported in late 2023 that even the top 25% of privately held startups have an average EBITDA of minus 30%. The Wall Street Journal, Financial Times and others are now confirming these risks. Scott Stanford, a cofounder and partner at ACME Capital, recently said that, “It’s not crazy to think half the VC firms that were actively investing in the last decade will be sidelined and eventually collapse.” Sara Ledterman, managing partner of 3+ Ventures, is even more pessimistic. She “estimates 3,200 funds have gone dormant over the past 14 months” while another “analysis found 8,538 VCs stopped investing in the past two years, leading to what some VCs have warned will be a “zombie reckoning in 2024.” Startup shutdowns rose at an accelerated pace in 2023. Pitchbook reports that around 3,200 startups failed in 2023 and more than 2,000 VC firms effectively halted making new investments in startups in the first nine months of 2023. Similarly, Carta reports that among users of its database, startup shutdowns rose to a record high of 264 in the first quarter of 2024 from 72 in the first quarter of 2022 . For publicly traded unicorns, eight went bankrupt or were acquired for below valuation prices in 2023 (Quotient technology, Coupa, Pear Therapeutics, Qualtrics, Blue Apron, Signify Health, Smile Direct Club, and Lordstown Motors). Several unicorns have followed suit so far this year: 23andme; Invitae; Nanostring Technologies and Fisker). Meanwhile, almost 400 tech companies have laid off workers this year so far, according to layoffs.fyi data — including two publicly traded unicorns (Gingko bioworks and Cue health). Endowments, pension funds, trusts and their peers are the limited partners (LPs) that are at the heart of the venture capital system, since they channel funds to venture capital firms. Business Insider recently reported that a growing number of limited partners have begun reneging on their commitments. LPs are legally liable for money they have promised the VCs, but few VCs want the bad press and possible domino effects from suing LPs. Better to pretend that all is for the best in this best of all possible worlds. Keep collecting fees and keep hoping that greater fools will soon appear.

A new crop of artificial intelligence startups has shaken up Silicon Valley — and the wider business world — throughout this year, but there’s one thing that hasn’t changed: Big Tech still wields power.

In the wake of Microsoft Corp.’s $10 billion investment in OpenAI in January, other tech giants raced to partner with leading AI startups through funding and cloud computing deals. Salesforce Inc. led a round in Hugging Face at a $4.5 billion valuation. Alphabet Inc. and Amazon.com Inc. invested billions in OpenAI rival Anthropic. And Nvidia Corp. seemed to back almost every AI startup of note.

This study introduces a comprehensive integrative framework that synthesizes the key determinants of collaboration between start-ups and large companies. This framework consolidates the prominent antecedents, configurations, and outcomes of such collaborations identified in previous research. By summarizing these factors, our framework serves as a valuable tool for comprehending and analyzing the dynamics of start-up and large company collaboration.

Regarding the antecedents of collaboration between start-ups and large companies, our findings align with the literature, emphasizing the importance of leveraging complementary resources and capabilities through collaboration (Dyer & Singh, 1998; Mesquita et al., 2017; Salvato et al., 2017). The divergent characteristics of firm size, organizational culture, and other factors render start-ups and large companies ideal partners in accessing untapped resources and capitalizing on synergies (Prashantham & Kumar, 2011; Allmendinger & Berger, 2020; Minshall et al., 2010).

A recent paper by two Chinese academics suggests that China is now widely recognized as a formidable contender in the realm of emerging and disruptive technologies (EDTs), particularly in the field of artificial intelligence (AI).

The communist nation, according to the researchers, has established itself as a global frontrunner in the practical implementation of AI across various sectors, including facial recognition, speech technology, financial technology (fintech), research publications, patents, and deep learning. Notably, China's advancements in AI hold significant implications for military applications, including nuclear warfare.