Sam Altman, the chief executive of OpenAI, has indicated that the leading AI companies will strike a collaborative agreement on safety, even as they compete fiercely for the same market.
Altman made the remarks in an interview for Fortune's series Fortune 500: Titans and Disruptors of Industry, published on Friday.
The magazine's editor-in-chief, Alyson Shontell, asked why he did not get in a room with Dario Amodei of Anthropic, Elon Musk of SpaceX and Demis Hassabis of Google DeepMind to agree a plan for the risks the technology poses.
"I think that will happen," Altman said.
He declined to say more about any talks under way.
"I'm not going to pre-announce private discussions that I think should be at some point shared as a group," he added. "But yeah, I think that will happen."
Responsibility over profit
Altman framed the case for coordination as a matter of duty rather than commercial advantage.
He said the industry's leaders carried a tremendous amount of responsibility, and could not let egos or the incentive for profit get in the way.
He told Fortune that OpenAI put safety above business considerations, and that a 10% chance of a catastrophic outcome was not acceptable.
He said the company's most advanced and still unreleased models were powerful enough that more work on safety was needed before pushing further.
Progress on capability, in his account, should not outrun the ability to monitor a model, to understand what it is doing, and to ensure it follows human values and its users' intent.
A week of alignment
The comments cap a week in which the industry's tone shifted towards caution.
Altman told Fortune that OpenAI would not hold a stock market listing in 2026, calling it an ill-advised moment given the state of AI safety.
Earlier in the week he told staff the company was considering slowing development of its most advanced systems, and might coordinate a pause with other labs, though he noted it was unclear whether all would agree, according to Bloomberg.
On Saturday Amodei published an essay urging the industry to slow the pace of improvement and committed Anthropic to giving independent evaluators permanent, employee-level access to its systems.
The backdrop is a public safety alarm.
Jacob Coxon, a former Anthropic researcher, resigned last Tuesday and accused both Anthropic and OpenAI of racing towards recursive superintelligence.
Evan Hubinger, Anthropic's alignment science lead, wrote on X that researchers earnestly believe AI could kill all humans, and put the risk within the decade at more than 10%.
The obstacle
Any formal pact faces a legal problem.
For rival companies to coordinate on commercial activity, such as the pace of model releases, risks breaching antitrust rules.
Amodei has argued the answer is government involvement, with officials in the room to confirm the talks amount to safety work rather than collusion.
Whether the leaders convene, and on what terms, remains the open question.
Altman's answer suggests the will is there, though he stopped short of naming a date.
