What Do AI Founders Think About Regulation? Tech Week Reveals a Divide Over Self-Policing

How are AI founders responding to the Trump administration’s call for the industry to police itself? At San Francisco Tech Week, founders and investors interviewed by The Associated Press welcomed a lighter approach to AI regulation, particularly as they looked for opportunities in defense and security technology. But those attendees do not represent every AI company or researcher. The question dividing the industry is whether companies’ own safety checks are enough when they also have strong incentives to build and sell new systems quickly.

Why did some Tech Week attendees welcome fewer AI rules?

At an event during Tech Week, held in San Francisco from October 5 to 11, 2026, founders and funders praised the administration’s support for AI development. AP reported that some expected a hands-off approach to improve prospects for businesses building AI-powered defense and security systems. The event was part of a wider week of gatherings for startups and investors, not a survey of the industry.

The appeal is straightforward: companies seeking funding or government customers may see fewer regulatory hurdles as a way to develop products and reach markets faster. Supporters of a light-touch approach also argue that extensive rules could slow U.S. companies competing internationally. Those are arguments about potential benefits, however, not proof that voluntary safeguards will prevent harm.

What does the White House’s self-policing agreement do?

On September 29, President Donald Trump hosted AI executives at the White House for a ceremonial signing of a voluntary safety agreement. According to AP, the brief pact calls on advanced AI labs to follow safety protocols and establish internal teams to monitor their work. Representatives of OpenAI, Anthropic, Meta, Google, SpaceXAI and Nvidia signed it.

Voluntary does not mean companies have promised to do nothing. OpenAI told AP it was working with outside safety assessors; Anthropic said it was bringing in independent evaluators. But the White House had not explained how it would monitor compliance with the pact when AP asked. An internal safety team also operates within a company whose business depends on developing AI. Outside evaluation can provide another check, but questions remain about what evaluators may examine, what they must disclose and what happens if they identify a serious risk.

The administration’s preference for industry-led safeguards does not mean there is no government activity on AI. Trump has established a federal AI task force, and AP reported that the Federal Trade Commission opened an investigation into potential consumer dangers involving AI companies. State and local officials are also considering or pursuing their own measures. An investigation, though, is not the same thing as an enforceable safety standard for every advanced AI system.

Do all AI founders oppose regulation?

No. The favorable reaction at Tech Week sits alongside calls for more oversight from people who have worked inside AI companies. David Robinson, a former OpenAI safety leader, told AP that a voluntary pact cannot replace law. At an October 5 New York City Council hearing, former industry employees warned about safety risks, while representatives of major AI companies described their own safeguards. Anthropic and OpenAI have also said advanced AI needs regulation.

Nor is there only one kind of AI rule under debate. Proposals can address testing before a system is released, reporting serious incidents, protecting consumers or giving independent reviewers access to safety information. People can favor some of those measures while opposing others they consider too broad or difficult to apply. Asking whether founders support “regulation” can obscure those differences.

Why does the distinction matter to everyone else?

AI companies have incentives to find risks: failures can damage their products, customers and reputations. They also have incentives to move fast when competitors are racing to release new capabilities. Self-policing leaves much of the decision about whether a system is ready, and how much the public learns about its risks, with the company that built it.

That trade-off matters beyond Silicon Valley. Concerns discussed in the current debate include AI agents interacting with outside systems, cybersecurity, consumer protection and the reliability of tools used in consequential settings. In an AP-NORC poll published October 8, 64% of U.S. adults said AI was developing too fast. The poll does not settle which rules would work, but it shows why enthusiasm at an investor event should not be mistaken for a public consensus.

The clearest answer is that AI founders do not hold a single view on regulation. Some Tech Week attendees welcomed the administration’s reliance on company-led safety efforts. Other industry voices want stronger external oversight. The unresolved policy question is not whether companies should test their products—they should—but who can verify those tests and require changes when a company’s judgment falls short.