Rare Truce Among Tech Giants: Sam Altman Agrees with Anthropic CEO on Need to Slow Frontier AI Development

OpenAI CEO Sam Altman has publicly backed an initiative by Anthropic CEO Dario Amodei calling for a controlled slowdown in the development of frontier artificial intelligence models. The fierce rivals agreed on a key point: the industry must intentionally slow the parameter race so that safety and alignment techniques can keep pace with the expanding capabilities of AI algorithms.

Amodei’s Manifesto: Bioweapon Risks and Loss of Control

The public discussion was sparked by Dario Amodei’s policy essay We Must Pace the Frontier. Anthropic’s founder argued that model capabilities are evolving faster than foundational research into their interpretability and reliability. Among the top threats in the coming years, he cited autonomous systems spiraling out of control, devastating automated cyberattacks, and the potential for algorithms to assist in developing biological weapons.

Amodei clarified that he is not advocating an outright ban on the technology. Instead, his proposal centers on “pacing”: deliberately slowing the marginal growth of frontier model capabilities. According to the researcher, buying even one or two years would give scientists time to develop workable mathematical safety guarantees before the first critically dangerous systems emerge.

The Anthropic CEO’s plan includes three practical steps:

1. Granting independent auditors access to internal development on par with full-time staff engineers to continuously analyze models during training;
2. Establishing a cartel agreement among developers to voluntarily cap runaway cluster scaling;
3. Setting up international intergovernmental oversight of the highest-risk areas of research.

Altman Responds, Opens Doors to Auditors

Altman voiced his support on X, confirming that pacing AI progress has become the central topic within OpenAI’s leadership in recent weeks.

“The idea of bringing in independent experts with access comparable to internal employees is great. We plan to do the exact same thing and will soon share details about the new audit format,” the OpenAI CEO wrote.

As Reuters notes, the aligned stance between the leaders of the world’s two top AI labs is a rare display of solidarity in the tech sector. Until now, fierce commercial rivalry has forced startups to rush releases out of fear of falling behind competitors.

The key question now is whether the companies can turn social media posts into a real self-regulatory mechanism without running afoul of antitrust laws or investor agreements.