Google Scrubs Pledge Not to Use AI for Weapons or Surveillance

Google this week deleted from its website a pledge not to develop artificial intelligence for weapons or surveillance. The change, first spotted by Bloomberg, wipes a section titled "applications we will not pursue" from the company's public AI principles page — a section that was still up as recently as last week.

This isn't a minor edit. That section was the explicit line Google had drawn for itself around uses of its technology it considered unacceptable. By removing it, the company no longer commits in writing to steering clear of military applications or mass surveillance tools.

Where That Pledge Came From

The commitment traces back to a 2018 clash between Silicon Valley and the Pentagon, when Google's involvement in a defense contract sparked internal employee protests. The company responded by publishing a set of AI principles that, from then on, included a public section ruling out the development of technology for weapons or surveillance systems.

That document became a touchstone within the industry: proof that one of the world's biggest tech companies was willing to walk away from certain business on ethical grounds. Its disappearance closes that chapter.

What Google Says Now

Asked for comment by TechCrunch, the company pointed to a new blog post on "responsible AI." It notes, in part, "we believe that companies, governments, and organizations sharing these values should work together to create AI that protects people, promotes global growth, and supports national security."

Google's newly updated AI principles state the company will work to "mitigate unintended or harmful outcomes and avoid unfair bias," as well as align itself with "widely accepted principles of international law and human rights."

The contrast is stark. Where there used to be a list of red lines, there is now an appeal to work alongside governments and an explicit nod to "national security." The language has shifted from prohibition to general principle: "we won't do this" has become "we'll do this carefully."

A Shift That Goes Beyond Google

The move fits a broader pattern. In recent years, much of the U.S. tech industry has drawn closer to the defense sector, following a stretch when military contracts were politically awkward internally. "National security" rhetoric has become commonplace in corporate messaging from major AI companies.

Google is no stranger to that friction. Its contracts to provide the U.S. and Israeli militaries with cloud services have sparked internal protests from employees in recent years. The company has maintained that its AI is not used to harm humans.

That defense, however, sits awkwardly next to what military officials themselves acknowledge. The Pentagon's AI chief recently told TechCrunch that some companies' AI models are speeding up the U.S. military's kill chain. That's the gap between the corporate message — "our AI does no harm" — and how customers actually put it to use.

Why It Matters

The significance here isn't that Google is about to start manufacturing weapons. It's what it means to walk back a voluntary public commitment.

These principles were self-imposed: nothing forced Google to adopt them, and nothing stops the company from erasing them now. That's precisely the weakness of self-regulation in AI. When the rules depend on a company's own goodwill, they can be rewritten the moment the business context shifts. What was a red line last week is no longer one this week — and updating a webpage is all it takes.

For anyone who argues the industry can police itself without outside rules, this episode makes the opposite case. Commitments that can be deleted at no cost and with no oversight offer little in the way of guarantees. Google's new language invokes "human rights" and "international law," but it swaps a concrete, verifiable prohibition for a far harder-to-check promise of good intentions.

What to Watch Next

The next signal won't come from another statement of principles — it will come from contracts. Without the "applications we will not pursue" section, there's no longer a public standard employees, clients, and outside observers can use to judge whether Google is crossing its own limits.

Internal pressure, which has already surfaced in protests over existing agreements, will remain one of the few real checks inside the company. And the deeper question — who decides, and with how much transparency, how far a tech giant's AI can go in the military domain — is now left without the answer Google once gave itself.