Congress stalls on military AI as Google and the Pentagon strike deal
Congress is far from passing military AI guardrails as Google and the Pentagon sign a new contract more permissive than for rival companies like OpenAI.
TL;DR
- Congress is lagging in establishing legal restrictions for military AI, despite growing concerns.
- The Pentagon has reached an agreement with Google to use its Gemini AI model for "all lawful use," including classified settings.
- The Google contract is reportedly more permissive than OpenAI's, with Google agreeing to adjust safety settings at the government's request.
- Critics, like Alex Turner of DeepMind, argue that Google's agreement lacks legal restrictions and relies on "aspirational language."
- Both Google and OpenAI state they have red lines against AI for autonomous weapons or mass surveillance, but these are seen by critics as non-legally binding commitments.
- Anthropic is the only major AI lab that has not yet made a deal with the Pentagon.
- Advocates are pushing for legislative safeguards, including greater transparency, modernized testing and verification, and a "five-second rule" for meaningful human control over AI weapons.
- Congressional committees are expected to mark up the National Defense Authorization Act this summer, where AI safeguards might be addressed.