Dario Amodei on use of Autonomous Weapons

Anthropic CEO and co-founder Dario Amodei delivered stark testimony before a U.S. Senate subcommittee, stating that artificial intelligence sophisticated enough to power autonomous weapons systems could be developed within the next two years. This warning places the timeline for a new era of AI-enabled warfare much sooner than many public estimates, directly challenging policymakers to establish regulatory frameworks before the technology becomes operational. Amodei, whose company builds the Claude AI models, framed this not as a distant sci-fi scenario but as an imminent policy challenge, urging for immediate action on international treaties and export controls to govern dual-use AI systems.

Amodei's testimony specifically highlighted how current frontier AI models, including those designed for benign purposes, contain capabilities that could be repurposed for lethal autonomous weapons (LAWS) with minimal modification. He argued that the same underlying architectures that enable advanced reasoning and planning for commercial applications could be adapted for targeting and engagement decisions in military contexts. This has sparked intense debate about whether companies like Anthropic, OpenAI, and Google DeepMind should restrict certain capabilities in their models or if governance must come entirely from state-level policy. The call for treaties mirrors historical non-proliferation efforts for chemical and nuclear weapons, but faces unique challenges due to the software-based, easily replicable nature of AI.