Anthropic CEO Dario Amodei has proposed pacing frontier-AI capability development so that safety, alignment, evaluation and interpretability work can catch up. In an essay published in September 2026, Amodei wrote that pacing would not mean halting model training or technical progress. That is his proposal and rationale; the reviewed evidence does not establish that slowing development is necessary, effective or sufficient.
Amodei also wrote that Anthropic was unilaterally committing to embedded external evaluators with ongoing, employee-like access to review safety practices and report incidents. The essay presents that commitment as an initial step toward the broader approach he advocates.
The proposal drew favorable public responses from several prominent AI leaders. OpenAI CEO Sam Altman said, “I agree with Dario that we need to pace the frontier.” Google DeepMind and Alphabet chief scientist Demis Hassabis said the essay pointed in the right direction, while adding that its details still needed to be worked out. Elon Musk responded to the essay by saying, “Dario is right.”
Those statements establish public endorsements or favorable responses to Amodei’s proposal. They do not establish a binding agreement among the companies, an implemented industry-wide slowdown or a change in any company’s deployment policy. Whether the leaders share the same meaning of “pacing” and oversight, and whether any of them will implement the proposed measures, remained unresolved.