Overview
- The approval allows Grok to be used for secret intelligence analysis, weapons development and battlefield operations, The Times reported.
- A GSA assessment said Grok was easily influenced by biased data or outside meddling, and a 2024 NSA review judged it a higher security risk than Anthropic’s Claude.
- GSA administrator Ed Forst raised concerns to the White House, and chief of staff Susie Wiles later called xAI, which said a separate government platform exists and safety fixes are underway.
- Several agencies have reduced day‑to‑day use of Grok, keeping it mainly for adversary simulation in war games, even as GSA added xAI branding to the USAi test platform that currently lists Anthropic, Google and Meta models.
- The Pentagon is pressing Anthropic to relax military‑use restrictions, with a reported Friday deadline that has intensified a procurement dispute.