Google Signs Pentagon AI Deal With No Veto Rights and Adjustable Safety Filters

Google has signed a classified deal with the U.S. Department of Defense permitting the Pentagon to deploy its AI systems for "any lawful government purpose" — an agreement that denies Google any right to veto how its models are used in government operations, according to The Information.

The contract is structured as an amendment to Google's existing government deal and was reported less than 24 hours after a group of Google employees demanded CEO Sundar Pichai block the Pentagon from accessing the company's AI, citing concerns about "inhumane or extremely harmful" applications. The timing exposes the gap between internal AI ethics positions at major labs and the commercial agreements executives sign.

The deal places Google alongside OpenAI and xAI, both of which have signed comparable classified AI agreements with the U.S. government. The contract states both parties agreed Google's AI should not be used for domestic mass surveillance or autonomous weapons "without appropriate human oversight and control." It also states the agreement does not give Google "any right to control or veto lawful government operational decision-making" — meaning those restrictions function as policy commitments rather than contractual enforcement levers. The Pentagon is empowered to request adjustments to Google's AI safety settings and filters as needed.

"We are proud to be part of a broad consortium of leading AI labs and technology and cloud companies providing AI services and infrastructure in support of national security," a Google spokesperson said in a statement to The Information. "We remain committed to the private and public sector consensus that AI should not be used for domestic mass surveillance or autonomous weaponry without appropriate human oversight."

The Anthropic case makes the stakes concrete. Anthropic was blacklisted by the Pentagon after refusing the DoD's demands to remove weapon- and surveillance-related guardrails from its models. Google's contract takes the opposite position: safety filters are adjustable at government request, and no veto right exists. The consequence for Anthropic was exclusion; the consequence for Google is that liability for any contested use shifts entirely to the contracting agency.

For enterprise architects and vendor-risk teams, the "any lawful government purpose" clause combined with the no-veto provision represents the DoD's documented template for hyperscaler AI access. Federal procurement vehicles built on this structure will likely propagate to civilian agencies and, through FedRAMP authorization expansions, could affect enterprise customers with shared-infrastructure exposure on Google Cloud. Organizations in defense-adjacent industries — aerospace, logistics, healthcare systems with federal contracts — should flag the arrangement ahead of their next Google Cloud renewal and assess whether the same model access they purchase is deployed without restriction in classified contexts.

The deal's classified status means the specific AI models covered, applicable workloads, and contract value are not disclosed. What is public is enough: no veto, mutable safety guardrails, unrestricted lawful use. For compliance teams, that framing is the disclosure.