Sam Altman stood before a crowd of business leaders at a Salesforce conference in San Francisco on Tuesday and delivered a message that split the difference between reassurance and alarm. The world has every reason to fear artificial intelligence, he said, but should trust the companies building it to police themselves.
"The world should trust that we are going to do the right thing because it's the right thing and we feel the magnitude of this," Altman told attendees.
The remarks marked Altman's first public appearance since a departing Anthropic researcher posted a warning last week that unchecked AI could kill all humans by the end of the decade. Several other AI executives and experts publicly backed the assessment, though without detailing how such a scenario would play out or what specific mechanisms could cause it.
That post triggered a fresh wave of scrutiny over the pace of AI development. Anthropic CEO Dario Amodei responded by calling on the entire industry to slow down and urging governments to step in with regulation. Altman, Google DeepMind co-founder Demis Hassabis, and Elon Musk all publicly endorsed Amodei's position.
But by Tuesday, the tone from industry leaders had shifted toward self-governance. Altman expressed confidence that OpenAI and its peers could regulate themselves without external oversight. "We will get it right, I'm very confident in our company's and industry's ability to do this safely," he said. He added that the industry would keep alignment and safety ahead of capabilities, and if that proved impossible, would "slow down or stop."
Mark Zuckerberg weighed in on X, writing that every AI lab has both the ability and the incentive to pursue safety-focused development independently. "Any lab that doesn't focus on alignment will fall behind," Zuckerberg wrote, pointing to significant liability exposure as a motivating factor.
Jensen Huang, CEO of Nvidia, reinforced the self-regulation stance at the same conference. The chipmaker, now the world's most valuable company by market capitalization thanks to surging demand for its AI hardware, has no interest in seeing the government slow the rollout of AI tools. "We don't need new laws or regulations," Huang said, dismissing what he called a false choice between speed of innovation and product safety.
Huang characterized safety as fundamentally an engineering problem. He said any AI company leader who loses confidence in their product should simply choose not to release it. "Run as fast as you can, but if at any time you feel the institution is not in control, take a pause," he said.
The pushback against outside regulation has not gone unanswered. Jack Clark, an Anthropic co-founder and executive, told the BBC that treating AI as a totally unregulated industry amounted to "rolling dice with immense risks." Patrick Hillman, chief operating officer of Logical Intelligence and an associate of AI luminary Yann LeCun, pointed to polling that shows Americans trust Silicon Valley less than Washington. "If you believe what you are building is dangerous, show us what you are prepared to stop doing," Hillman said.
Some industry figures have argued that the resurgence of AI fears is being amplified to generate hype, rather than reflecting genuine existential risk.
Industry Groups Push Voluntary Safety Frameworks
Leaders from both OpenAI and Anthropic indicated they have begun exploring industry-wide safety agreements in recent weeks. Amodei made a brief appearance at the Salesforce conference and said Anthropic is now in dialogue with the rest of the industry about committing to stronger safety standards and more rigorous checks on AI tools and development pipelines.
"We didn't appreciate that it would lead to companies growing so fast, and how quickly they would become central to things," Amodei said of the broader impact of the AI boom.
Chris Lehane, an OpenAI executive, said last week that the company is collaborating with Anthropic and Google DeepMind to advance frontier AI standards through a voluntary effort that would proceed regardless of whether governments provide support. "With stakes this high, we cannot let the perfect become the enemy of the good," Lehane wrote.
Altman used his conference appearance to pitch AI tools as both a productivity aid and a defensive measure against AI-powered cyber attacks, urging business leaders to adopt the technology rather than retreat from it.
The split between those calling for government intervention and those advocating industry self-policing reflects a deeper tension in the AI sector. Companies like OpenAI and Anthropic have invested heavily in alignment research and safety testing, but critics argue that voluntary commitments carry no enforcement mechanism. The industry's track record on self-regulation in areas like data privacy and content moderation does not inspire confidence among skeptics.
As AI capabilities advance at a pace that consistently outstrips predictions, the question of who gets to decide when enough is enough remains unresolved. For now, the companies building these systems say they should be trusted to make those calls themselves.