FTC efforts fail at stopping AI monopolization---it’s too late because they already have control.

At these pivotal moments we often see anti-competitive conduct as the monopolist attempts to protect their powerful position. Where Facebook chose to buy, Microsoft attempted to bury. They used multiple strategies to exclude Netscape from the market, resulting in the landmark antitrust case against them. Unfortunately, today’s Big Tech platforms are even more powerful than past monopolists. They saw this transition coming and have been preparing for years. In some ways it appears they have already completely transitioned their existing gatekeeper control to the new industry. When Sam Altman and OpenAI wanted to take their AI products to the next level, Altman knew he would need immense computing power. Although he had initially hoped to keep his company in a not-for-profit model, he recognized how critical it was to get access to that computing power. As a result, he agreed to a partnership with Microsoft. Google’s mechanism for locking up search has been exclusive default contracts with browsers, mobile device manufacturers, and others; the subject of last year’s antitrust trial against them that has yet to be decided. In the trial, Microsoft’s Satya Nadela was asked how the new technology of generative AI was changing or would change the search market. His answer was that it didn’t change the fundamentals of the market. The default contracts continue to prevent Microsoft and Apple, the two largest potential competitors, from competing effectively against Google. But the story is not over. As we speak, AI companies are still in a mad dash to gain access to as much content as possible to train their models. AI companies are licensing content, scraping it from the web, or leveraging data that they already control. This is not finished and it is happening today. But control of data is only one piece of the puzzle—there are other features of the AI industry that will make maintaining dynamic competition extremely difficult including a small pool of skilled researchers and developers and intensive capital investment costs for training supercomputers and the cloud computing infrastructure necessary for offering AI services at scale. As a result, even with pro-competition policies it is likely that we will see significant consolidation around existing technology gatekeepers that already have access to the inputs necessary to succeed in the AI industry. Even more so than the platform markets that dominate our world today, AI may be a market that will tend towards tipping to monopoly or oligopoly. Facing down the prospect that the AI ecosystem may be dominated by a few powerful companies, that makes it all the more critical to decisively regulate AI to protect people from real harms like racial bias and discrimination, labor exploitation, disinformation, invasions of privacy and autonomy. At the same time, I don’t want to give up on competition. That means applying fair competition rules to prevent AI gatekeepers from leveraging their power in one part of the stack into others. This might include utility-style regulation for cloud infrastructure to ensure fair access. It should also mean robust public sector investments in lowering barriers to entry into AI, through research and infrastructure and building full-stack public AI options that can compete with private AI models.