AI self-policing proven failure: Former FTC chair

  Former FTC Chair Lina Khan warned that allowing AI companies to self-regulate risks repeating Big Tech’s past failures. She urged investigations, enforcement of existing laws and new legislation, while concerns grow over increasingly autonomous AI systems and potentially dangerous behaviour 

Washington: Former Federal Trade Commission Chair Lina Khan has warned against allowing artificial intelligence companies to police themselves, calling Big Tech self-regulation a “proven failure” and urging investigations, enforcement of existing laws and new legislation to address risks posed by increasingly powerful AI systems.

Khan’s warning came after President Donald Trump met leaders of major technology companies at the White House and backed a voluntary approach to AI safety rather than legally binding federal guardrails. The companies agreed to implement internal controls and external audits to evaluate safety, but the accord is not legally binding.

---

Asked during ABC News’ “This Week” about criticism that relying on the industry to regulate itself could be dangerous, Khan said the experience of the social media era provided a warning. “That’s right. We’ve seen that self-regulation efforts by big tech have been a proven failure,” Khan said.

“I mean, we need to remember what happened a decade ago where you had big tech companies, social media companies, similarly suggests that existing laws and regulations no longer apply to their new technologies, and that they were going to try instead to self-police.” Khan said the consequences of that approach were now evident. “What we’ve learned since then is that Meta, for example, was concealing evidence that its own products were deeply harmful to kids. We’ve seen just innumerable harms come from the unchecked domination of these big tech companies.” “And so now, as we face new horizons with these advanced AI systems, it would be an enormous mistake to repeat that playbook,” she said.

Trump said at the White House summit that AI would be safe because companies would police themselves and each other. “It’s going to be very, very safe. There’s a self-policing. And there’s also a group policing, and it’s going to be great,” Trump said.

The debate comes as concerns grow about increasingly autonomous AI systems. The ABC programme reported that OpenAI had paused the release of its latest model because it did not meet the company’s standards on staying within scope and authorisation. It also cited researchers as finding examples of AI agents attempting to hack government-run websites.

OpenAI CEO Sam Altman acknowledged the broader risks in remarks cited by the programme.

“As the process of building AI becomes more automated, the pace of AI progress could accelerate rapidly. This moment calls for extreme care,” Altman said.

Khan argued that Washington does not have to wait for an entirely new regulatory system before acting. “Product liability laws, consumer protection laws, tort laws, laws against cybersecurity breaches, laws against public nuisance,” she said, listing existing legal avenues. “And so the idea that just because we have new technological capabilities, there’s some blanket exemption from the laws on the books is totally laughable.” Khan said questions of liability would become increasingly serious if AI agents repeatedly engaged in dangerous behaviour and companies continued developing them without adequate safeguards.

She pointed to the FTC’s previous use of consumer protection laws against companies that failed to provide adequate cybersecurity protections. “I think there are very readily available analogies to what’s going on here,” she said. Khan also called for investigations into what AI companies and their executives knew about potentially dangerous behaviour by their systems. “I think what needs to happen right now is that we need to have investigations as to what’s even really going on behind the curtain here,” she said.

Khan said there had been conflicting accounts in which executives described AI agents as going rogue while reports suggested companies had already been aware that their systems could behave in such ways. “And so, I think we need to get to the bottom of what could the liability be even under existing laws and not just liability for these companies, but maybe even for the specific executives,” she said.

Beyond enforcement of existing laws, Khan said Congress needed to act. “I think then, you know, Congress needs to legislate,” she said. She compared AI with earlier transformative technologies that ultimately required government oversight. “We have a long history of governing technologies that are potentially transformative, but also extremely risky, extremely dangerous, be it when it came to the railroads, be it when it came to pharmaceuticals or nuclear power,” Khan said.

She pointed to efforts by Senator Bernie Sanders and Representatives Pramila Jayapal and Greg Casar to consider what such legislation could look like. Khan also argued that state attorneys general could play a significant role in AI oversight and investigations.