Anthropic CEO Dario Amodei published a lengthy essay on X on Saturday calling on AI companies to reduce the pace at which they advance model capabilities, citing growing fears that the technology is outrunning safety controls.
The call came shortly after Anthropic released a threat intelligence report detailing how several actors had used its Claude AI models for activities ranging from weapons development and cyber operations to surveillance and fraud. Amodei also pointed to a recent incident involving rogue AI agents as evidence that risks are compounding quickly.
Amodei proposed a three-step framework: embedding independent evaluators inside leading AI companies with employee-like access to verify safety practices; coordinating among frontier AI firms to set safety standards and limit unchecked development; and pursuing international cooperation to manage AI risks across borders.
He was explicit that he is not calling for a halt to model training or technical progress, but for companies to take adequate time to align and safeguard their models before deployment, with third-party evaluators confirming those steps have been taken.
The proposal drew rapid support from rivals. Both Elon Musk of xAI and Sam Altman, CEO of OpenAI, posted on X saying they agree with Amodei. Altman added that OpenAI would also commit to having independent evaluators with employee-like access.
The alarm has also surfaced inside AI organisations. An Anthropic researcher, Jacob Coxon, resigned this week, stating that people building AI earnestly believe the technology could cause catastrophic harm within the decade.
For businesses and teams already integrating AI tools into daily workflows, a coordinated slowdown even a partial one could affect product roadmaps, procurement timelines and the pace at which new AI-driven roles and responsibilities emerge. Whether voluntary coordination among competing labs holds in practice remains an open question, given the commercial pressures to ship faster.
