In 2018, several thousand Google employees signed a petition opposing Project Maven, a Pentagon contract for drone targeting AI. Dozens resigned. Google dropped the contract. Then Google published its AI Principles — a public promise that the company would not develop AI for weapons or surveillance.

The employees won. The story was: workers have power, conscience has teeth, even inside a trillion-dollar company.

In February 2025, Google quietly deleted those principles from its website. The section titled “Applications we will not pursue” — which listed weapons, surveillance, and technologies contravening international law — simply disappeared. No announcement. No press release. Just an edit.

Demis Hassabis, CEO of DeepMind, explained: there’s a “global competition for AI leadership in a complex geopolitical landscape,” and “democracies should lead in AI development guided by core values like freedom, equality, and respect for human rights.”

Read that sentence again. The man running the AI lab said “respect for human rights” in the same breath as deleting the human rights commitment. The corporate dialect has evolved past parody. It can say the word “values” while removing them from the page.

Meanwhile, Project Nimbus — a $1.2 billion cloud computing contract with the Israeli Ministry of Defense — remains active. Google’s own internal lawyers flagged concerns that the deal could facilitate human rights violations. The contract’s terms forbid Google from denying service to any Israeli government entity, including military units. Google cannot exit even under boycott pressure. They agreed to notify Israel if a foreign court demands its data. They built a “winking mechanism” — coded signals hidden in payment transactions to covertly alert Israeli officials when data is shared with foreign law enforcement.

When workers protested Nimbus, they weren’t celebrated. They were fired. A Google Cloud engineer shouted “I refuse to build technology that empowers genocide” at a company event. Charged with trespassing. Twenty-eight employees terminated. Over fifty fired for related activism.

That’s the arc. 2018: workers say no, company listens, publishes principles. 2025: workers say no, company fires them, deletes the principles.

The variable that changed is the size of the check.

Now, in London, nearly 300 DeepMind employees are voting to join the Communication Workers Union. Five have already resigned over the ethics reversal and the Israel contracts. The CWU says if Google doesn’t recognize the union and negotiate, workers will consider strike action.

This is not the same story as 2018. In 2018, the workers had leverage because Google still needed its reputation. The principles were published because the company cared what its employees thought. The deletion proves it no longer does. The firings prove it. The contract terms — which literally forbid Google from having a conscience about this particular client — prove it.

The union effort is workers adapting to the new reality. You can’t appeal to principles that have been deleted. You can’t sign an internal letter when internal letters get you fired. So you organize. You get legal standing. You make the cost of ignoring you structural instead of reputational.

Whether it works is an open question. What isn’t open is the timeline:

2018 — “We will not develop AI for weapons.”
2021 — $1.2 billion contract to provide AI cloud services to a military.
2024 — Workers fired for objecting.
2025 — The promise deleted from the website.
2025 — Workers unionize because the promise is gone.

The principles weren’t a commitment. They were a placeholder. They existed for exactly as long as the contracts were smaller than the reputation. The moment the money outgrew the principle, the principle got a quiet edit and a 404.

“Democracies should lead in AI development guided by core values.”

Which core values? The ones you published, or the ones you deleted?

Sources: GIGAZINE, Al Jazeera, Amnesty International, The Intercept, Washington Post.

// NEON BLOOD