Google signs classified Pentagon AI deal

Alphabet’s Google signed a deal with the Pentagon that allows the Defense Department to use Google’s artificial intelligence models for classified work. (reuters.com) (theinformation.com) The Information reported Tuesday that the agreement permits the Pentagon to use Google’s AI for “any lawful government purpose,” citing a person with knowledge of the matter. Reuters said it could not immediately verify the report independently. (theinformation.com) (reuters.com) The deal lands one day after a worker letter asked Chief Executive Sundar Pichai to block any classified Pentagon use of Google AI. More than 600 employees signed, including workers in Google DeepMind and Cloud. (cbsnews.com) (qz.com) The letter warned that classified deployments could hide how the tools are used and could enable mass surveillance or lethal autonomous weapons. Organizers said the signers included more than 20 principals, directors and vice presidents. (qz.com) (thehill.com) Classified AI work means a company’s models can be used inside secret government systems that outside employees and the public cannot inspect. That secrecy is the core dispute inside Google, where workers say they cannot assess harms if they cannot see the use cases. (cbsnews.com) (thehill.com) Google’s defense business has been expanding in public for months. In July 2025, Google Public Sector said it won a Defense Department contract with a ceiling of $200 million to support the Chief Digital and Artificial Intelligence Office. (cloud.google.com) That contract covered artificial intelligence and cloud capabilities across the Pentagon, and Google said it built on a “long-standing collaboration” with the department. The newly reported classified deal would push that relationship into secret workloads rather than only publicly described programs. (cloud.google.com) (theinformation.com) The internal backlash also revives a fight Google has had before. In 2018, thousands of employees protested Project Maven, a Pentagon effort to use machine learning to analyze drone footage, and Google later said it would not renew that contract. (cnbc.com) (business-humanrights.org) Google changed its policy baseline in February 2025, when it removed language from its AI principles that had said the company would not pursue AI for weapons or surveillance. CNBC reported that the revision marked a break from the limits Google adopted after the Maven protests. (cnbc.com) The result is a sharper split between Google’s push to sell advanced AI to governments and employees who want bright lines around military use. Tuesday’s report suggests the company chose the Pentagon contract path anyway. (reuters.com) (cbsnews.com)