Anthropic CEO warns AI lacks judgment, rejects autonomous weapons contracts

Anthropic, the AI safety company behind the Claude model, has drawn a hard line on military AI. CEO Dario Amodei formally declined US Department of War contracts for fully autonomous weapons, stating that AI models cannot be relied upon to exercise the critical judgment comparable to trained military personnel. In his January 2026 essay “The Adolescence of Technology,” Amodei characterized the current AI phase as a pivotal moment loaded with geopolitical risk, flagging dangers including propaganda manipulation, strategic misuse by bad actors, and the risk of misaligned AI applications spiraling beyond their intended purpose. Anthropic also incorporated costly measures to prevent its models from assisting in bioweapon development.

In April 2026, Anthropic’s Mythos model identified tens of thousands of software vulnerabilities across existing systems. Amodei warned that governments and organizations have roughly 6 to 12 months to reinforce their cyber defenses before competing models, including those developed in China, catch up. While Amodei has not referenced cryptocurrencies in his warnings, the cybersecurity implications are significant: the same vulnerability-hunting capability applies to smart contracts, wallet infrastructure, and exchange platforms. The cybersecurity sector stands to benefit most directly, as organizations rush to patch vulnerabilities within that critical window.