Google’s decision to let the Pentagon use its AI models for classified work has exposed a sharp divide over how far Big Tech should go in serving national security when the technology’s risks are still poorly understood.

What the deal does

Both liberal and conservative accounts agree on the core terms: Google will provide AI models to the Pentagon for “any lawful government purpose,” joining OpenAI and Elon Musk’s xAI in offering tools for classified use.¹² The contract explicitly says Google has no right to “control or veto lawful government operational decision-making,” meaning the military, not the company, ultimately decides how the systems are used.²

Google insists this is a responsible, limited role. The company says it is part of a “broad consortium” supporting national security, applying AI to areas such as logistics, cybersecurity, translation, and critical infrastructure defense, and that offering API access to its commercial models on its own infrastructure “represents a responsible approach.”² The agreement also includes language that the AI system is “not intended for, and should not be used for, domestic mass surveillance or autonomous weapons … without appropriate human oversight and control,” a position Google publicly reiterates.¹

Liberal worries vs. conservative backing

Liberal coverage emphasizes the risks and internal backlash. The Guardian highlights “significant disagreements with the Pentagon and major employee pushback” over similar deals, noting that classified networks handle “mission planning and weapons targeting” and that Google must help adjust safety settings at the government’s request.¹ Critics argue that once access is granted, corporate ethics become largely symbolic.

Conservative reporting, by contrast, foregrounds national security needs and frames Google as “proud to be part” of the defense effort.² It notes the Pentagon has sought similar “any lawful” deals across the AI sector and points to Anthropic’s penalties for resisting such broad terms, casting Google’s cooperation as pragmatic rather than reckless.²

Similarities and differences

Both perspectives acknowledge intense employee opposition, including a letter from more than 600 staff warning that classified work risks enabling “lethal autonomous weapons and mass surveillance.”² Both also stress that the contract’s guardrails are non-binding on government operations.

Where they diverge is in the balance of fear and trust: liberal outlets highlight the potential for mission creep and opaque militarization of AI, while conservative coverage stresses deterrence, resilience, and confidence that “lawful” use and human oversight are sufficient checks. Beneath the rhetoric, however, both sides describe a structural reality: the Pentagon is methodically locking in access to commercial AI, and once inside classified networks, the true use of these systems will largely escape public scrutiny.

---

1. Google Reportedly Signs Classified AI Deal with US Pentagon — Report on Google’s classified AI deal, scope of Pentagon access, and employee pushback.

2. Google Agrees to Deal with Pentagon for AI Usage in Classified Projects — Details on "any lawful government purpose" terms, Google’s national security justification, and internal employee letter.