Google DeepMind VP Tom Lue confirms the company is “leaning into” military contracts after scrubbing anti-weapons pledges from its 2025 AI principles.
For years, Google’s relationship with the military was a source of internal shame. The company effectively pinky-swore to avoid “weapons of war” after the 2018 Project Maven protests. But that era of Silicon Valley pacifism is officially over.
At a recent town hall, Google DeepMind VP Tom Lue dropped the pretense.
He reminded employees that the company’s AI principles were quietly updated in 2025, scrubbed of specific pledges against surveillance and weapons development. The new metric for taking a government contract is now remarkably flexible: whether the “benefits substantially exceed the risks.”
It isn’t just a change in wording; it is a change in the company’s soul.
While rivals like Anthropic are currently tied up in federal court for refusing to drop ethical “red lines” regarding autonomous weaponry, Google is leaning in. DeepMind CEO Demis Hassabis even noted he is “very comfortable”- working with democratic governments is a path to global safety.
The logic is simple.
The Pentagon is currently rolling out “Gemini for Government” to three million personnel, and Google wants a seat at that table. By framing the work as “administrative” or “clerical,” Google provides itself a layer of plausible deniability. Yet, the removal of the surveillance ban suggests the ceiling for this partnership is much higher than a glorified secretary.
Google’s “North Star” used to be its “Don’t Be Evil” manifesto.
Now, it mimics a calculated cost-benefit analysis. As the line between civilian tech and national security blurs, Google has decided that being a “supply chain risk” is a far greater danger to its bottom line than a few disgruntled employees.
