Google Signs Pentagon AI Deal with No Veto Power, Igniting 580-Employee Revolt

More than 580 Google employees, including 20-plus directors, VPs, and senior DeepMind researchers, signed an open letter to CEO Sundar Pichai demanding the company refuse classified Pentagon AI contracts. According to The Washington Post, the letter specifically targets use of Google's Gemini models on air-gapped military networks where the company cannot monitor how its AI is deployed. This new protest, reported by Bloomberg and Business Insider, marks a significant escalation from the 3,100-employee revolt in 2018 over Project Maven, the drone-footage analysis program that ultimately led Google to decline renewal of that contract.

On 28 April 2026 The Verge reported that Google and the Pentagon have already signed a classified agreement granting the Defense Department broad rights to use Gemini for “any lawful” military purpose. Crucially, the contract strips Google of any veto power once the model is running inside classified networks. Employees learned of the deal’s final terms only after the letter had circulated, turning an internal warning shot into open revolt. The company had not disclosed the agreement’s lack of usage restrictions during earlier all-hands briefings, according to two people present.

The current protest centers on classified environments where traditional oversight becomes impossible. Once Gemini operates on classified military networks, Google loses visibility into whether its AI is being used for autonomous weapons or mass surveillance. This differs from Project Maven's publicly documented use for drone video analysis. Employees argue that “trust us” becomes the only remaining safeguard when systems operate in military black boxes. The letter specifically references Google's removal of its own AI weapons red lines last year, which the company had established following the 2018 protests.

The Department of Defense has significantly expanded its AI initiatives since Project Maven, with classified programs now seeking advanced capabilities like Google's Gemini. The Washington Post notes this represents a shift from basic image recognition to more sophisticated AI applications in military contexts. The Pentagon's approach has become more aggressive following tensions with China and Russia, pushing for rapid deployment of frontier models in secure environments. The new contract reportedly includes provisions for continuous model updates delivered via encrypted channels, ensuring the military always has access to the latest capabilities.

Senior leadership has responded with a mix of reassurance and silence. Pichai told staff the company retains “contractual assurances” the tech will comply with international law, but has not explained how those assurances are enforced without visibility. Some VPs privately complain the deal undermines years of ethical AI branding, while others argue commercial realities require defense revenue. The internal chat rooms lit up when The Verge’s story dropped, with one senior engineer writing, “We just handed the keys to the kingdom and promised not to look.” Another posted a screenshot showing the contract’s “any lawful use” language in red highlight.

The employee coalition is preparing a second, sharper letter demanding immediate suspension of the contract and a public ethics board review. Sources say at least 200 more signatures arrived within hours of The Verge’s report. Legal teams inside Google are quietly studying whether the contract’s lack of usage restrictions could expose the company to future liability if the AI is implicated in civilian harm. Meanwhile, the Pentagon is already requesting integration timelines for Gemini into classified targeting systems, according to procurement documents viewed by The Verge.