Google expands Pentagon AI access on classified networks

Google has agreed to give the U.S. Department of Defense expanded access to its artificial intelligence tools for use on classified networks, under terms that allow “all lawful applications,” according to a report dated April 28, 2026.

The arrangement makes Google the third AI company to secure a deal with the DoD on these terms, following similar agreements reached by OpenAI and xAI, officials and reporting said.

DoD deal follows Anthropic dispute over unrestricted terms

The move comes after AI model maker Anthropic declined to provide the Pentagon with comparable unrestricted access, seeking to add guardrails aimed at preventing its systems from being used for domestic mass surveillance and autonomous weapons, according to the report.

After Anthropic’s refusal, the Pentagon labeled the company a “supply-chain risk,” a designation that triggered a legal fight. Last month, Anthropic obtained an injunction against the DoD’s designation, the report said.

Contract language and enforceability questions

Google’s agreement reportedly includes language opposing the same types of uses Anthropic sought to restrict. However, the report said questions remain about how enforceable those provisions would be in practice.

The report described the DoD access as expanded and tied to classified networks, indicating the tools would be available in environments that handle sensitive information, while still being limited to lawful uses.

Employee pushback highlights internal debate

The expanded access also arrives amid internal pressure at Google. An open letter signed by 950 Google employees urged the company to adopt Anthropic’s more restrictive approach to military AI contracts, the report said.

The report did not specify whether the employee letter led to changes in the final terms, or how Google plans to monitor compliance with any restrictions referenced in the agreement.