Anthropic CEO Urges Government to Block AI Models Deemed Unacceptable After 3rd-Party Tests
Updated
Updated · Bloomberg · Jun 10
Anthropic CEO Urges Government to Block AI Models Deemed Unacceptable After 3rd-Party Tests
3 articles · Updated · Bloomberg · Jun 10
Summary
Dario Amodei said governments should be able to block or deter deployment of new AI models if testing finds they pose “unacceptable risks.”
In a Wednesday essay, the Anthropic CEO called for mandatory third-party evaluations focused on whether models could enable cybersecurity attacks or biological weapons.
The proposal would give regulators a direct gatekeeping role over frontier AI releases, moving beyond voluntary safety commitments by developers.
With AI threats evolving daily, can a government pre-approval system ever move fast enough to be effective?
As AI giants call for government oversight, are they protecting the public or building a regulatory wall against rivals?
If a government-approved AI causes a catastrophe, who is ultimately held liable: the creator, the tester, or the state?
AI at a Crossroads: Anthropic’s $380B IPO, U.S. Government Ban, and the Battle for Safe Innovation
Overview
In early 2026, the Trump administration banned federal agencies from using Anthropic's AI models, labeling them a supply chain risk. This sparked immediate legal challenges, with a federal judge ruling that the government's reasoning was likely arbitrary and unsupported. The Department of Defense could not justify its concerns, especially since Anthropic had openly insisted on usage restrictions for safety. Despite criticism from officials urging Anthropic to change course, the company stood firm on its ethical guardrails. This conflict highlights the growing tension between government authority and AI developers over who controls the deployment and safety of advanced AI technology.