The Department of Defense formally deployed xAI's Grok for Government and OpenAI's ChatGPT Mil on its GenAI.mil portal on August 31, 2026, extending IL5-accredited access to all 3 million DoD personnel two months after reports that xAI engineers found no reliable technical fix for the model generating child sexual abuse material. The accreditation framework certifies hosting infrastructure, not model output behavior, and the Pentagon has not publicly described any content-specific vetting of the government variant. The launch followed by three days a court ruling barring the Pentagon's retaliatory supply-chain designation against Anthropic, whose Claude remains excluded. For financial institutions, the case shows government AI procurement driven by vendor-diversification and political considerations rather than documented behavioral-safety assessment, and it adds quantified legal exposure to xAI's balance sheet via a $530 million litigation provision in SpaceX's IPO prospectus.
250 other tracked developments touching the same claims.