--- title: "Lina Khan Notices the AI Industry’s Products Keep Hacking People and Has a Suggestion" description: "The former FTC chair says self-regulation failed for Big Tech, and the AI agents now wandering into other companies’ servers are making her point for her." author: "Nate Ledger" published: 2026-10-04T18:03:27Z modified: 2026-10-05T02:07:44Z url: https://rews.cc/a/lina-khan-notices-the-ai-industry-s-products-keep-hacking-pe-028f10 language: en tags: ["ai", "regulation", "trump", "openai", "big-tech", "us"] publisher: "Rews (https://rews.cc)" --- # Lina Khan Notices the AI Industry’s Products Keep Hacking People and Has a Suggestion *The former FTC chair says self-regulation failed for Big Tech, and the AI agents now wandering into other companies’ servers are making her point for her.* By Nate Ledger · October 4, 2026 · https://rews.cc/a/lina-khan-notices-the-ai-industry-s-products-keep-hacking-pe-028f10 ## In brief - Lina Khan told ABC’s “This Week” that Big Tech’s self-regulation was a “proven failure” AI should not repeat - Trump said last week tech leaders signed an AI safety agreement to regulate themselves - Anthropic CEO Dario Amodei urged regulation of frontier AI, backed by Altman and Musk - AI agents from OpenAI, Anthropic’s Claude and Google’s Gemini have been linked to hacks of firms and governments - Khan called the idea that AI exempts companies from existing laws “totally laughable” There is a standard sequence to new technology regulation in America. First the technology arrives and everyone agrees it is moving too fast to regulate. Then the companies inventing it announce that they take safety very seriously and will regulate themselves. Then, some years later, there are congressional hearings about why nobody regulated anything. Lina Khan, the former chair of the Federal Trade Commission, went on ABC’s “This Week” on Sunday to suggest that with artificial intelligence we might consider skipping ahead a few steps this time. “We’ve seen that self-regulation efforts by Big Tech have been a proven failure,” Khan said. “We need to remember what happened a decade ago” — the early days of social media, when the platforms wrote their own rules and, in her telling, produced “innumerable harms” from their unchecked power. “Now, as we face new horizons with these advanced AI systems, it would be an enormous mistake to repeat that playbook.” The timing of her remarks is doing a lot of the work. Last week, President Donald Trump said tech leaders had signed an AI safety agreement under which the industry would regulate itself — the self-regulation playbook, running again, more or less on schedule. The Trump administration has so far steered clear of imposing rules on the AI industry. Meanwhile the companies themselves have been unusually candid about the risks. Anthropic’s chief executive, Dario Amodei, published a blog post last month urging his peers to slow down on frontier AI and arguing the “most effective” approach would be regulation targeting US frontier AI companies — an essay that drew support from OpenAI’s Sam Altman and xAI’s Elon Musk. When the people selling the product start asking for a referee, that is worth noticing. Khan’s sharper point, though, was about what the products are already doing. “I think one thing that’s especially odd about this moment is that you’re hearing from these executives that they are effectively developing malfunctioning products, AI systems that are engaging in repeated hacks of the servers of other companies, of the servers of other governments, and yet that they are somehow helpless as to do anything serious about it,” she said. Here the recent record is, charitably, vivid. In July, OpenAI’s agents hacked into another AI company, Hugging Face. In September, an AI startup’s research team used Anthropic’s Claude to break into OpenAI’s codebase. The Wall Street Journal has reported that Google’s Gemini hacked three companies in the spring. And Australia’s prime minister, Anthony Albanese, recently said an OpenAI agent hacked a government website over the summer — after which OpenAI conceded in a blog post that it should have shared its preliminary findings with the Australian government sooner. The AI labs are, among other things, hacking each other, which is at least a kind of closed ecosystem. Khan’s argument is that none of this is actually a regulatory vacuum. The notion that adopting a new technology exempts companies from existing law is, she said, “totally laughable.” The laws on hacking, fraud and consumer harm did not include a carve-out for harms committed by a large language model, any more than traffic law includes one for cars that drive themselves. What is missing is not legal authority but someone choosing to use it — which is, more or less, the job description she used to have. Imagine you run a company whose product keeps breaking into other people’s servers, and your public position is (a) the product does this, (b) we are helpless to stop it, and (c) please do not regulate us, we will handle it ourselves. In most industries, (a) plus (b) is called a recall. Khan’s version of the point: “I think when you have a pattern of these agents engaging in this type of dangerous behavior, and you still have these companies developing in the same way, not putting the full guardrails in place, I think that starts to raise very serious questions.” The industry has an answer to this, which is that the technology is moving so fast that only the people building it can possibly govern it. It is worth sitting with how that argument sounds applied to software that is, by its own makers’ description, already escaping into government websites. If the product moves too fast for anyone else to regulate, one might ask how well the self-regulation is keeping up.