Anthropic CEO Proposes Slowing AI Development for Safety

Anthropic CEO Dario Amodei has called for the artificial intelligence industry to deliberately slow its pace of model development to allow time for implementing critical safety measures, according to details first reported by The Hill.

In a blog post published Saturday, Amodei outlined what he termed a "pacing framework" that would require AI companies to take "adequate time to align and safeguard their models" before releasing new capabilities. The proposal comes amid a fresh wave of warnings from current and former employees at major AI companies about potential risks from the technology.

The three-part framework

Amodei's approach centers on three core elements. First, he proposed giving external evaluators "employee-like access" to AI companies, enabling them to oversee safety implementations and report incidents directly. Anthropic committed to implementing this step immediately.

Second, he called for AI companies and democratic governments worldwide to introduce comprehensive safety procedures. Third, he urged democratic nations to pressure authoritarian governments, specifically naming China, to agree to prohibitions on "obviously dangerous uses" of the technology.

The CEO emphasized that pacing does not mean halting model training or technical progress entirely. "Progress will still seem fast, and we must make wise use of the time we gain," Amodei wrote.

Why it matters

The proposal represents a significant shift from an AI industry leader, acknowledging that the current development pace may be outstripping safety capabilities. With Anthropic competing directly against OpenAI, Google, and other major players, a call to slow down from within the industry carries particular weight. The timing also suggests mounting internal concern about AI risks, coming the same week multiple employees issued public warnings.

Policy and legislative response

Sarah Heck, Anthropic's head of policy, endorsed the proposal and expressed interest in working with Congress to implement the suggested guardrails. She specifically called for government action including blocking advanced chip sales to adversarial nations and enacting national laws requiring frontier model testing.

Rep. Don Beyer, a Virginia Democrat who co-chairs the House AI Caucus, wrote Wednesday that "doing nothing on AI is unacceptable and dangerous." He urged lawmakers to hold hearings, demand answers from companies, and pass ready legislation.

Amodei concluded his post by reaffirming his belief in AI's potential to "enormously improve the quality of human life," while stressing that benefits will only materialize "if we build the technology in the right way." He acknowledged the measures would not be easy to implement but argued "we owe it to humanity to try."

The Hill first reported these details on September 12, 2026.