Why Google DeepMind researchers are resigning over military AI

Google employees are quitting because their AI is going to war. And this time, Google doesn't seem particularly interested in stopping.

From Project Maven in 2018 to a Pentagon deal in 2026

In 2018, thousands of Google employees protested Project Maven, a Pentagon program using AI to analyze drone footage. Google eventually decided not to renew the contract.

Eight years later, we're watching the sequel. Except management appears to have changed the ending.

In April 2026, hundreds of employees urged CEO Sundar Pichai not to let Google's AI models operate in classified US military systems. Google went ahead with a Pentagon deal anyway.

Why researchers are leaving

Now researchers at Google DeepMind are resigning.

Sofia Liguori, a London-based research engineer who spent three years at Google, told the BBC she felt “ashamed” of the company's growing involvement in military AI.

Andreas Kirsch, another researcher who resigned in October, questioned whether Google's governance and oversight were sufficient for deploying AI in classified military environments.

Their concern is pretty specific. The Pentagon can use these systems for classified work, while employees worry Google has limited visibility into what happens after deployment and little power to veto particular uses.

That's a rather uncomfortable arrangement when the technology can assist intelligence analysis and military decision-making.

And it gets better.

In February 2025, Google revised its AI principles, removing its explicit commitment not to develop AI for weapons or certain surveillance uses. Apparently, the ethical guardrails were due for a software update.

Google says it reviews contracts against its AI principles and acceptable-use policies, and provides employees with channels to raise concerns without retaliation.

There is a legitimate argument for military AI. Better intelligence, defensive cybersecurity and logistics can save lives. But how much human oversight is enough when an AI-generated recommendation might influence who gets targeted?

And who's accountable when the model confidently gets something wrong?

The lesson for CTOs and AI leaders

For CTOs and AI leaders, there's another lesson here. If you hire researchers to build responsible AI, you need to be prepared for them to ask what “responsible” means when your biggest customers come calling.

In 2018, Google employees helped change a corporate decision.

In 2026, some are realizing that the only decision they can change is where they work.

Turns out “Don't be evil” gets a lot more complicated when the customer has aircraft carriers.

If you too are looking for your new role, my team can take care of your entire job search, from applications to direct hiring manager outreach while you prepare for interviews.