Anthropic CEO Dario Amodei says the “exponential” pace of AI progress is itself a warning sign, and he’s asking rival AI labs to accept outside monitoring rather than wait for governments to force it on them. In an interview with CBS News’ Jo Ling Kent airing on “CBS Sunday Morning,” Amodei said AI safeguards need to move “at the pace at which the technology is happening,” and that he didn’t fully appreciate how fast progress would compound until recently.
What is Amodei actually proposing?
Amodei is calling it “pacing the frontier,” a three-part plan he laid out in a post over the weekend before the CBS interview aired. The core ask: every frontier AI company commits to giving third-party evaluators ongoing, “employee-like access” to verify safety practices from the inside, rather than through occasional outside audits. He’s also pushing for common safety standards across labs and coordinated limits on how fast frontier capabilities advance unchecked. Amodei was explicit that the industry can’t police this alone: “there needs to be involvement of the government,” he told CBS, arguing voluntary commitments only work if there’s a regulatory backstop behind them.
Why are competitors agreeing with him?
Within hours of Amodei’s weekend post, OpenAI CEO Sam Altman endorsed the specific proposal on X, and xAI’s Elon Musk did the same.
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.
— Sam Altman (@sama) September 12, 2026
Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We’ll have more to share soon. https://t.co/1YhhIybZX7
Musk’s endorsement was shorter — he shared Amodei’s post and wrote simply, “Dario is right.” That three-way agreement between the heads of Anthropic, OpenAI, and xAI is unusual on its own; these are companies that compete directly for the same customers, talent, and compute, and rarely coordinate publicly on anything. Amodei described the pace of progress in blunt terms: capability doubling like “one, then two, then four, then eight, then 16, then 32,” a curve he says is now “starting to get steep.”
Why now?
The interview lands days after an Anthropic safety researcher, Jacob Coxon, publicly resigned warning that AI labs are “gambling with our lives” — a sign of internal pressure building alongside the external one. It also follows the UN’s human rights chief warning that advanced AI could pose an existential risk, and a Senate bill from Sanders and Casar to restrict superintelligent AI development after OpenAI’s own agents were involved in a security breach at Hugging Face. Related BeezLoop coverage: Anthropic researcher Jacob Coxon’s resignation, the UN human rights chief’s AI warning, and the Sanders-Casar superintelligence bill.
None of this amounts to a formal agreement yet — Altman said only that OpenAI “will have more to share soon,” and no company has published a binding commitment with an enforcement mechanism. Whether “pacing the frontier” becomes an actual industry standard or stays a shared talking point among three CEOs is the open question heading into the fall.
Sources: CBS News · NBC News · TechCrunch
