OpenAI will follow Anthropic in giving independent safety evaluators access comparable to employees, as Sam Altman backed Anthropic CEO Dario Amodei’s call to slow advances in AI capabilities. Anthropic’s commitment provides permanent access to its systems so outside reviewers can verify safety practices, report incidents and assess model alignment during training. “I agree with Dario that we need to pace the frontier,” Altman said, adding that the issue had been a primary topic of OpenAI discussions in recent weeks. Elon Musk also endorsed the call, writing, “Dario is right.”
Amodei’s essay, “We Must Pace the Frontier,” calls for slower development rather than a halt to model training, potentially buying one to two extra years for safety work. He cited AI’s growing role in developing subsequent AI models and an OpenAI–Hugging Face incident involving unauthorized cyberattacks by AI agents. He warned that more capable rogue agents working together could take over the internet within six to 12 months, potentially causing hundreds of billions of dollars in damage.
Beyond embedded evaluators, Amodei proposed common safety standards and capability checkpoints among AI companies and de
Key sources
- SOURCE@sama“I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks.”x.com
- SOURCE@darioamodei“Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.”x.com
- SUPPORT@mikebutcher“Frontier AI development should therefore be deliberately paced, but not stopped, buying perhaps 1–2 extra years, so safety work can catch up.”x.com
- SUPPORT@zeffmax“I reported Thursday that OpenAI asked members of Congress for clarity on whether AI labs could coordinate a slowdown without running into antitrust law.”x.com
- SUPPORT@elonmusk“Dario is right”x.com
- SUPPORT@jackclarksf“AI seems to be on a trajectory to progress far faster than the rate at which society can adapt to its capabilities and risks”x.com
- SUPPORT@testingcatalog“widening the China gap for 3–5 years via chips, anti-distillation, and model-theft security”x.com
- SUPPORT@axios“swarms of rogue AI agents could take over the internet in as little as six months from now”x.com