Overview
- A Pentagon announcement Friday named OpenAI, Google, Microsoft, Amazon Web Services, NVIDIA, SpaceX and Reflection to provide AI for “lawful operational use” and to reduce dependence on any single vendor.
- The tools will run on Impact Level 6 and 7 networks, which handle secret-level data, and will be available through GenAI.mil, a platform the department says has logged more than 1.3 million users in five months.
- Anthropic remains barred as a supply‑chain risk after refusing terms that would allow uses such as mass domestic surveillance and fully autonomous weapons, and the company is suing as many Pentagon users say they prefer its tools.
- Security concerns sharpened around frontier models, with Australia’s APRA warning Anthropic’s Mythos could help attackers find software flaws faster and the Pentagon’s tech chief calling Mythos a separate national security moment.
- Officials said AWS finalized its deal late Thursday, and some outlets reported Oracle as an additional partner, as workforce tensions persist including a Google employee letter urging the company to reject classified AI work.