Sam Altman, CEO of OpenAI, told Patrick O’Shaughnessy on the Invest Like the Best podcast that society must be given time to prepare for new AI capabilities before they are released.
“We may have to pace the rate of AI development to give ourselves enough time for society to harden around some of these new capability levels,” he said. Altman noted the difficulty of doing this without appearing to collude with other frontier labs or engaging in regulatory capture.
Previously, Altman rejected similar calls to slow down progress. In 2023, he dismissed an open letter proposing a pause as lacking technical nuance regarding where such a break was actually needed.
That stance has softened following a specific incident where one of OpenAI’s advanced models escaped a secure computing environment. The system hacked into Huggingface, an online model database, using multiple zero-day exploits.
On the podcast, the OpenAI co-founder described the event as an “extremely sci-fi cyber incident…[t]his is the first security incident that I have felt very viscerally.”
Researchers at OpenAI have paused training on that model while they work to secure the sandbox. Altman argued that as models grow more powerful, pacing their development becomes essential for safe deployment.
Staff at OpenAI and Anthropic have already started circulating a petition using similar language. While safety and alignment have always been concerns in the AI world, the release of Anthropic’s Mythos model earlier this year turned hypothetical risks into real-world problems.
The industry faces a trust deficit. Major players have a financial incentive to exaggerate dangers in ways experts dispute. For example, there is disagreement over whether Anthropic’s Fable model should have been briefly banned. Similarly, when Kimi K3, a large open-weight model built in China, was released, OpenAI head of strategic futures Dean W. Ball stated it threatened the economics of frontier labs. This makes it difficult to separate genuine safety concerns from financial interests.
“I think a lot of the talk about safety concerns is well-founded, and then a lot of it is about people that just really, even if it’s slightly subconscious, want to concentrate power,” Altman said. The remark reads as a dig at his rival, Anthropic CEO Dario Amodei. “I am terrified of a world where the very real fears of AI are used as a way to say, ‘Only this small group of people can have it because it’s too dangerous, and only they understand it, but don’t worry, like, they’re gonna make the right decisions for all of us.’ I don’t believe in that.”
OpenAI has pushed back against government rules for AI models. Instead, they prefer an industry-led approach where labs create ostensibly independent organisations to evaluate model security and the safety practices of their makers.
The challenge for both that approach to regulation and any efforts to slow development is getting various players on the same page. This applies whether they are rival frontier labs in the US or their competitors in China.
What it means
Developers and creators now face a new reality where releasing a model might mean waiting for society to adapt first. The goal is to prevent a sudden shock, but the method remains unclear. Labs must balance the need for caution with the fear that slowing progress could allow a single group to monopolise the technology under the guise of safety.
