NCCapital

Google Pentagon AI Deal Sparks Military AI Concerns Google has reportedly signed a classified agreement with the U.S. Department of Defense to provide access to its AI models for sensitive government work, including classified applications. The deal marks a major shift for a company that once stepped away from Pentagon AI work after internal backlash over Project Maven in 2018. According to reports citing The Information, the agreement allows Google’s AI systems to be used for “any lawful government purpose.” That language has drawn scrutiny because it appears broad enough to cover military planning, intelligence work, and other classified uses where outside oversight may be limited. The deal comes shortly after more than 600 Google employees urged CEO Sundar Pichai to reject classified Pentagon AI work. Their letter warned that classified use could make it difficult to know whether the company’s technology was being used in ways that violate Google’s stated commitments around responsible AI. The backlash echoes Google’s 2018 Project Maven controversy. At that time, thousands of employees protested the company’s role in a Pentagon program that used AI to analyze drone footage. Google eventually declined to renew that contract and later published AI principles limiting work related to weapons and surveillance. The new reported agreement suggests Google is now taking a more permissive approach to national security AI contracts. The contract reportedly includes language stating that Google’s AI is not intended for domestic mass surveillance or autonomous weapons without human oversight. However, reports say Google does not have veto power over lawful government operations, raising questions about how enforceable those limits would be in practice. Another sensitive part of the deal is the reported requirement that Google help adjust AI safety settings and filters at the government’s request. That detail matters because AI safety filters are often what prevent models from producing certain types of harmful, restricted, or operationally sensitive outputs. If those safeguards can be modified for classified government use, critics may question how much control companies really retain once their systems enter military environments. Google is not alone. Reports say OpenAI, xAI, and other major AI companies have also pursued or signed defense-related agreements as the Pentagon pushes to integrate commercial AI tools into military and intelligence workflows. These deals reflect a larger shift: frontier AI companies are increasingly treating national security work as a major market rather than a line they will not cross. Anthropic has reportedly taken a more restrictive position on certain military use cases, including autonomous weapons and surveillance. Reuters reported that tensions over those restrictions led the Pentagon to label Anthropic a supply-chain risk, a designation the company has challenged. For Google, the central issue is trust. The company says its work supports national security and that it remains opposed to domestic surveillance and autonomous weapons without human oversight. But employee critics argue that classified contracts make those assurances difficult to verify. The broader question is not whether governments will use AI. They already are. The question is whether private AI companies can meaningfully enforce ethical limits once their models are deployed inside classified military systems. Google’s reported Pentagon deal shows how quickly the AI industry’s boundaries are changing. A company that once retreated from military AI work after employee pressure is now reportedly joining the Pentagon’s classified AI ecosystem. For workers, watchdogs, and policymakers, that shift will likely intensify the debate over who controls advanced AI — the companies that build it, or the governments that buy it.