Search the archive.

Roman V. Yampolskiy, "Human Indignity: From Legal AI Personhood to Selfish Memes," arXiv:1810.02724, 10-2-2018, https://arxiv.org/abs/1810.02724

If, on the other hand, AIs are going to become more intelligent than humans, the indignity for humanity would come from being relegated to an inferior place in the world, being outcompeted in the workplace and all other domains of human interest [23, 24]. AI led corporations would be in position to fire their human workers. This would possibly lead to deteriorating economic and living conditions, permanent unemployment and potentially reduction in rights, not to mention further worsening of the situation including to the level of existential catastrophe (extermination) [25].

Ryan Abbott and Alex Sarch, Punishing Artificial Intelligence: Legal Fiction or Science Fiction, 2019, https://lawreview.law.ucdavis.edu/issues/53/1/articles/files/53-1_Abbott_Sarch.pdf

Finally, and perhaps most worryingly, conferring legal personality on AI may lead to rights creep, or the tendency for an increasing number of rights to arise over time.229 Even if AIs are given few or no rights initially when they are first granted legal personhood, they may gradually acquire rights as time progresses. Granting legal personhood to AI may thus be an important step down a slippery slope. In a 1933 Supreme Court opinion, for instance, Justice Brandeis warned about rights creep, and argued that granting corporations an excess of rights could allow them to dominate the State.230 Eighty years after that decision, Justice Brandeis’ concerns were prescient in light of recent Supreme Court jurisprudence such as Citizens United v. Federal Election Commission and Burwell v. Hobby Lobby Stores, which significantly expanded the rights extended to corporations.231 Such rights, for corporations and AI, can restrict valuable human activities and freedoms.

Today , there is nothing in the law that explicitly prohibits computers as inventors. 164 If the USPTO or a federal district court does not find that a computer can be an inventor on a patent application under the current law, it will be challenging to maintain the ideology that patents "promote the progress of science." 165 In the event that computers meet the current statutory requirements for inventorship, there are two additional reasons why they should be granted inventorship on a patent application. First, corporations make up 93.7% of all patent assignees in the United States, and they are well equipped to protect computer-generated innovations via trade secret law if they remain unprotectable under patent law. 166Link to the text of the note Second, the Framers and the Supreme Court emphasized the "public good" purpose of the Constitutional patent provision and laws. 167 Accordingly, it is foreseeable that if the USPTO or a federal district court does not look past the statutory language that merely suggests that inventors must be human, then as AI becomes widespread, a large number of computer-generated innovations will be guarded from the public.

Law Explorer, 2015, “Intellectual Property and Corporate Value,” https://lawexplores.com/intellectual-property-and-corporate-value/

In 1974, the number of utility patents granted was 76,278. By 2004, the number of patents granted more than doubled, to 164,293. Corporations were granted the vast majority of patents in 2004; foreign and domestic corporations received eighty-eight percent of the total number of patents granted. Interestingly, forty-four percent of patents are owned by U.S. corporations and forty-four percent are owned by foreign companies. Foreign and U.S. individuals combined received ten percent of all patents, with the remainder going to U.S. and foreign governments. The 2004 distribution of patent ownership has remained largely unchanged for the past ten years (see ).

The lag between technological developments in artificial intelligence (AI) and the law that governs it is growing as the pace of AI adoption increases. In patent law, open questions have evolved from whether AI is itself patentable to questions surrounding inventorship where AI is involved in the development of an invention. Inventorship is a key inquiry in patent law, as an inventor owns the patent unless and until the inventor assigns the patent to another entity.[11]

AI, an invention itself, is now being used to develop new patentable inventions, notably in the pharmaceutical industry, where it is being used both to discover new drugs and repurpose existing drugs.[12] Two main questions about inventorship arise when AI is involved in the invention process: is AI itself an inventor under current patent law? And who is the proper inventor if not AI itself?

“Putting the Person in PHOSITA: The Human’s Obvious Role in the Artificial Intelligence Era” 2021, 62 B.C. L. Rev. 1413 (2021), https://lawdigitalcommons.bc.edu/bclr/vol62/iss4/8

New technologies present new challenges. As the patent system adapts to addressing the needs of expanding AI capability, it is important that past lessons inform modern practice. Even minor attempts to alter the doctrine based on whether a given technology contributes to an invention could throw the test into disarray. Current obviousness analysis provides a sufficiently flexible standard for evaluating inventions created by or with the aid of AI technology. Unless or until current obviousness doctrine proves unable to accommodate this expanding technology, we should not rush to alter it.

The analogous art test limits the scope of prior art to preexisting technologies within the “field of the endeavor” and to those technologies that are reasonably pertinent to the problem solved by the invention.232 This test recognizes the ingenuity of those who examine technology from one sphere of study and apply it to something completely different.233 At some point, the ability to draw inspiration from prior art in one practice area and apply it to solve a problem in a completely different area is nonobvious.234Arguably, AI is best suited to enhancing human capabilities in this particular area of invention.235 Unlike human inventors who sometimes have mental barriers and preconceptions that inhibit looking for inspiration far-afield, AI machines do not have such hang-ups.236 Rather, AI machines simply look for patterns within preexisting technology by assigning values to the data signals they receive.237 Not only will AI better equip humans to look beyond the blurry lines separating one field from the next, but it has the computing power to sift through an endless array of data for a solution to a problem.238

In 1966, in Graham v. John Deere Co., the Supreme Court imposed upon inventors the heightened burden of showing nonobviousness against the wider scope of prior art incorporated by new technologies.239 As a result, once AI is common in a given industry, inventors will have to meet the heightened burden of showing nonobviousness based on what a PHOSITA aided by AI—as well as any other widely available technology—would find reasonably pertinent to the problem solved by the invention.240 Further, in KSR, in 2007, the Supreme Court endowed the PHOSITA with the “inferences and creative steps” that an ordinary worker would exhibit.241 Such a PHOSITA would take advantage of the relevant technologies employed in the industry to solve routine problems.242 Thus, current doctrine accommodates the increasing role that AI plays in inventing by expanding the scope of prior art to include tools that a PHOSITA would have access to and by endowing the PHOSITA with the good sense to use them.243

A. The Current Obviousness Doctrine Will Reach the Correct Results for Inventions Created Using AI Technology

The thrust of the modern obviousness analysis focuses not on the inventor, but rather on the invention itself.220 The test hinges on whether a claimed invention presents a significant advancement over prior art.221 Therefore, the mere fact that an invention is created in part or entirely by an AI machine does not render it obvious under current doctrine.222 Section 103 of the Leahy-Smith AIA explicitly reaffirmed that the “manner in which [an] invention was made” will not support an obviousness determination.223 With or without the use of AI technology, the determination comes down to whether an average worker, with knowledge of all relevant, preexisting technology would find the claimed invention obvious.224 For example, an invention created by a team of researchers over the course of five years is not necessarily any less obvious than if it had been created by one researcher in a few days using an AI machine.225 Suppose that, in the above example, researchers in the inventor’s field widely use AI technology and that the researcher using the AI is a worker of ordinary skill.226 The researchers who labored away for five years should not receive a patent for something that would have been obvious within a few days had they applied the tools commonly used in the industry.227 Far from incentivizing innovation, allowing a patent in such a situation would reward the researchers for failing to stay up-to-date on the standard technology used in the field.228

Note: PHOSITA, pronounced “Fo-sit-ah,” means “Person Having Ordinary Skill In The Art”

[Edited for Grammar]

Allowing AI to become named authors or inventors of their derivative works despite major difficulties in properly identifying the original source of creation will create a lack of public trust in the copyright and patent systems due to the public’s inability to monitor the creative entity and track when the idea will reach the public domain. Thus, because the improper identification of authors and inventors creates an inability for the public to monitor privatization, and the inability of the public to monitor privatization creates a lack of public trust in the copyright and patent systems, we can conclude that the improper identification of authors will create a lack of public trust in the copyright and patent systems.

As previously discussed on IPKat, it is difficult to comprehend how an algorithm that is anything short of an artificial general intelligence (AGI), can be said to be capable of invention (IPKat: ). However, we are currently far away from anything that would approach an AGI. The capabilities of today's AI algorithms are restricted to narrow and well-defined tasks, as can be seen from a brief glance over Google/Deepmind's patent portfolio (see ). A major theme identified by the USPTO from the consultation results was that, based on the majority view that AGI has not yet arrived current AI could neither invent nor author without human intervention. As the AIPPI commented in their submissions to the consultation:

the current state of AI technology is not sufficiently advanced at this time and in the foreseeable future so as to completely exclude the role of a human inventor in the development of AI inventions.

Jane Croft, June 15 2022, “Inability to patent AI creations could hit business investment,” Financial Times, https://www.ft.com/content/eaff4b88-988c-409c-b0c6-86573dfcd253

Noam Shemtov, professor in IP and technology law at London’s Queen Mary University, notes that most AI experts hold the view that this threshold will only be reached in 2075, and so the current law is sufficient. “It therefore makes little sense for the patent regime to gear up at present for such speculative development,” he says.

Jyh-An Lee, “Computer-generated Works under the CDPA 1988,” Artificial Intelligence and Intellectual Property, 04-19-2021, https://global.oup.com/academic/product/artificial-intelligence-and-intellectual-property-9780198870944?cc=cr&lang=en&

However, this line of argument is not flawless. First, AI developers normally obtain copyright for their generative AI software. If the software meets the patentability requirements, it can be protected by patent law as well. Therefore, AI developers, incentivized by the existing IP regime, are already able to profit from their commercially successful software through licensing royalties.107 Additionally, software innovators’ competitive edge is reinforced by the first-mover advantage, which is typically significant in the software industry and provides another layer of incentive for software development.108 Therefore, if software developers can obtain additional copyright for AI-generated works, they would be over-rewarded for the same creation. This additional protection is not justified from a policy perspective.109 After all, it is the software, rather than the human author, that creates the work. Different from human authors, computer programs will not be incentivized to create more or fewer works by copyright.110 If software developers are entitled to the copyright of AI-generated works, they might own an indefinite number of subsequent copyrights by creating one single AI project, because of its continuously generative nature.111 IP protection and its consequent social costs will then become disproportionate to the incentives it provides.

Reto Hilty, Jörg Hoffmann, and Stefan Scheuerer, “Intellectual Property Justification for Artificial Intelligence,” Artificial Intelligence and Intellectual Property, 04-19-2021, https://global.oup.com/academic/product/artificial-intelligence-and-intellectual-property-9780198870944?cc=cr&lang=en&

The commonly known issue of underinvestment in the production of new knowledge has been traditionally addressed via either direct legal intervention in the marketplace, by granting subsidies for research and development or by transforming the public into a private good through the granting of intellectual property (IP) rights. Over the last decades, new technologies have boosted the general tendency of establishing subject matter-specific forms of protection, such as for computer programs or databases. This move towards a policy-driven activism has rightly been criticized.1 Indeed, the false assumption that IP rights a priori provide incentives for creation and innovation is not tenable. One also should bear in mind the potentially dysfunctional effects of excessive IP protection and related deadweight loss.

Reto Hilty, Jörg Hoffmann, and Stefan Scheuerer, “Intellectual Property Justification for Artificial Intelligence,” Artificial Intelligence and Intellectual Property, 04-19-2021, https://global.oup.com/academic/product/artificial-intelligence-and-intellectual-property-9780198870944?cc=cr&lang=en&

3.2.1.1 General incentive theory

[Barry C. Lynn is the executive director of the Open Markets Institute. Over the past two decades, Lynn pioneered understanding of how the monopolies of the 21st century threaten our democracy, individual liberties, security, and prosperity. Lynn’s efforts to update anti-monopoly law and thinking for the digital era have been fully embraced by the Biden administration and have shaped the thinking of policymakers and scholars around the world. His warnings on structural flaws in international systems predicted today’s supply chain crises, and his proposed remedies have been widely studied by the U.S. government, Europe, Asia, the IMF, and the OECD.]

For more than two decades, the authors and publishers of news, books, music, photography, and other forms of easily digitizable materials have fought with little success to protect their copyrighted properties and livelihoods from the power of the platform monopolists. This includes a long list of lawsuits against Google and other corporations for not preventing individuals and businesses from sharing copyrighted materials on platforms ranging from YouTube to Google News without permission. Such transmission of data all but eliminates the ability of copyright owners to get paid for their work, while also removing most of the incentives for consumers to pay for what they read, watch, and listen to. Now the rapid rollout of generative AI is taking this existing suite of problems and amplifying it at an almost astronomical rate, in four specific ways. The first new question posed by AI is how creators will be paid for the use of their materials in training foundation models and applications such as ChatGPT and search engines. The U.S. Copyright Office has launched an inquiry into these issues, to assess whether copyright laws need to be updated in response to AI. The IAC – one of the world’s largest internet holding companies and home to digital publishers such as DotDash Meredith – recently warned in its comment to the Copyright Office that if generative AI firms don’t pay publishers for copyrighted content, the technology will undercut not only the news media industry, but all kinds of other web publications that deliver reliable and safe information. A second new question posed by AI is whether it can be used to create art that until now has required human beings as producers – such as music, books, news, and films. For writers, this threat is so tangible that prohibiting the indiscriminate use of AI-generated text in screenwriting was a key demand during the five-month strike led this year by the Writers Guild of America, which successfully secured the guarantees it demanded. As the Open Markets Institute put it in a letter to the U.S. Copyright office: “Enforcing copyright laws and requiring companies to meet their legal obligations – whether that is compensating rights holders or ensuring transparency of training data – are critical in order to protect human creativity and reward the labor that goes into it.” A related set of questions arise over whether ideas and inventions generated by AI models should be granted patents, and if so, who gets the credit. These concerns are not just hypothetical. An analysis by the Washington Post of a major data set used by Google and Meta’s main LLMs found that the biggest single category of websites were patent, business, and industrial sites including Kickstarter and Patreon, raising concerns that the technology may copy businesses’ and creators’ original ideas in suggestions to users. Granting patent rights to AI models poses a wide variety of risks to the entire system of intellectual copyright. One potential risk is simply empowering large corporations to further reinforce their dominance. For example, a pharmaceutical corporation that owned the largest datasets on chemical structures and prescriptions could use AI to mass-generate patents for potential new drug compounds, driving market concentration and excluding rivals. Finally, generative AI is enabling dominant retail platforms like Amazon and music streaming services like Spotify to host and promote AI-generated books, music, and similar products. Writers are reporting that Amazon is recommending AI-generated rip-offs of their own published books on its marketplace. Meanwhile, the e-commerce giant refuses to disclose how many of these bogus publications it prevents from being published or how many it takes down. Spotify not only generates revenues from AI-generated music that violates copyright but hosts an undisclosed number of bots that act as listeners of both AI-generated song rip-offs and real artists – effectively reducing royalties for the latter.

Ryan Abbott, “The Reasonable Robot: Artificial Intelligence and the Law,” University of Surrey, 01-24-2020,

As discussed earlier, given that AI is functioning as an inventor, and likely inventing at an increasing rate, it would seem that the Patent Office should be receiving an increasing volume of applications claiming AI as inventors. That the Patent Office apparently has not suggests that applicants are choosing not to disclose the role of AI in the inventive process. Again, this may be due to legal uncertainties about whether an AI inventor would render an invention unpatentable. Without a legal inventor, new inventions may be ineligible for patent protection and may enter the public domain after being disclosed.

In the last few years, artificial intelligence (AI) has taken centre stage in public, academic, and political discussions. Generative AI, in particular, produces new content in response to prompts, offering transformative potential across multiple fields such as education, entertainment, healthcare, and scientific research. While generative AI is expanding quickly, it is also leading to several uncertainties in areas such as human rights, existing technologies (for example, ), and migration, amongst others. Terrorist groups are also growing interested in exploiting this new technology and using it to their advantage.  

The use of technology by terrorist groups is not a new topic, as technology has always played a vital role in the work done by terrorist organisations. However, with the latest developments in artificial intelligence and generative AI, some terrorist groups now consider the “benefits” of these technologies. The advantages of generative AI for malicious purposes are notable - from interactive recruitment to developing propaganda and influencing people’s behaviour via social media channels. New technological advances open multiple doors to infinite possibilities that terrorist groups are prepared to exploit and use to their advantage.  

Catastrophic AI risks can be grouped under four key categories which we explore below, and in greater depth in CAIS’ linked paper: Malicious use: People could intentionally harness powerful AIs to cause widespread harm. AI could be used to engineer new pandemics or for propaganda, censorship, and surveillance, or released to autonomously pursue harmful goals. To reduce these risks, we suggest improving biosecurity, restricting access to dangerous AI models, and holding AI developers liable for harms. AI race: Competition could push nations and corporations to rush AI development, relinquishing control to these systems. Conflicts could spiral out of control with autonomous weapons and AI-enabled cyberwarfare. Corporations will face incentives to automate human labor, potentially leading to mass unemployment and dependence on AI systems. As AI systems proliferate, evolutionary dynamics suggest they will become harder to control. We recommend safety regulations, international coordination, and public control of general-purpose AIs. Organizational risks: There are risks that organizations developing advanced AI cause catastrophic accidents, particularly if they prioritize profits over safety. AIs could be accidentally leaked to the public or stolen by malicious actors, and organizations could fail to properly invest in safety research. We suggest fostering a safety-oriented organizational culture and implementing rigorous audits, multi-layered risk defenses, and state-of-the-art information security. Rogue AIs: We risk losing control over AIs as they become more capable. AIs could optimize flawed objectives, drift from their original goals, become power-seeking, resist shutdown, and engage in deception. We suggest that AIs should not be deployed in high-risk settings, such as by autonomously pursuing open-ended goals or overseeing critical infrastructure, unless proven safe. We also recommend advancing AI safety research in areas such as adversarial robustness, model honesty, transparency, and removing undesired capabilities.

Neil Turkewitz, “Blurry Lines: The Complex Intersection of Race & Intellectual Property”, The Shadow, 16 February 2021, https://medium.com/the-shadow/blurry-lines-the-complex-intersection-of-race-intellectual-property-fe88e64d4058

This underscores that, as Professor Vats has observed, intellectual property standards are built on racial scripts, and that such gaps can — and frequently do — operate to deny ownership on the basis of race, or at least in a manner largely aligned with race. But the point I want to stress is that whether in the halls of WIPO, national capitals around the world seeking to address protection of folklore/traditional cultural expression, or as reflected in the Blurred Lines case, the relevant “racial script” that operates to the disadvantage of people of color relates to limitations on the scope of protection, and not to an excess thereof. This is not to suggest that issues of access and use do not pose special challenges for communities of color — only to highlight that an accurate narrative concerning race and intellectual property is complex, and involves, in large part, the failure to provide adequate protection to cultural productions.