Google has changed the way it frames acceptable uses of its artificial intelligence technology, removing earlier restrictions that had ruled out weapons and surveillance systems. The update signals a broader opening toward AI defense contracts at a time when several major AI companies are building or supplying systems for military, defense, or intelligence uses.
According to the Washington Post, the company has dropped its previous ban on using AI technology in weapons and surveillance systems. Google now says its guidelines will rely on human oversight and testing, with the goal of aligning work with accepted legal and human rights principles while reducing harmful outcomes.
What Google changed in its AI guidelines
The previous version of Google’s AI guidelines was more explicit about what the company would not build. It prohibited four categories of applications: weapons, surveillance, technologies that "cause or are likely to cause overall harm," and anything that violates international law and human rights.
Those restrictions mattered because they drew a clear line around certain uses of AI. Weapons and surveillance were not treated as ordinary product categories. They were named as areas where the company would not apply its technology.
The new guidelines take a different approach. Instead of listing the same prohibitions, they focus on oversight, testing, and alignment with "widely accepted principles of international law and human right" while "minimizing unintended or harmful outcomes." That language does not describe the same hard ban found in the old guidelines.
For Google, the change is not being presented as a retreat from ethics, but as a response to a more competitive global AI environment. The company says the shift reflects the growing global race for AI leadership.
We believe democracies should lead in AI development, guided by core values like freedom, equality, and respect for human rights
That statement was written by Google's Head of AI Demis Hassabis and Senior Vice President James Manyika in a blog post.
Why the shift is significant
The change is striking because Google’s earlier position grew out of intense internal pressure. In 2018, thousands of Google employees protested military contracts and argued that Google should not be in the business of war.
The new wording marks a clear departure from that earlier stance. The company is no longer keeping the same stated restrictions on weapons and surveillance systems in its AI guidelines. Instead, it is emphasizing governance, testing, and human oversight as the framework for deciding how AI should be used.
That change also affects how Google fits into the wider AI industry. The company is not moving in isolation. Other major AI companies have already entered partnerships or arrangements connected to defense contractors, the US military, or US intelligence and defense agencies.
Defense partnerships are becoming more common
The broader technology sector is increasingly active in defense-related AI work. OpenAI recently partnered with defense contractor Anduril to develop AI-powered drone defense systems for the US military.
Meta has given the US military access to its Llama AI models. Anthropic is working with Palantir to provide US intelligence and defense agencies specialized versions of Claude through Amazon Web Services.
Microsoft reportedly proposed using OpenAI's DALL-E image generator to develop military operations software for the US Department of Defense last year.
Taken together, these examples show a sector-wide shift. AI companies that once faced heavy scrutiny over military work are now becoming more willing to provide models, systems, or specialized versions of their products for defense-related use.
The new debate around AI defense work
The central issue is not only whether AI can be used in defense settings. It is also who decides what counts as acceptable use, what safeguards are enough, and how companies define responsibility when their technology moves into military or intelligence contexts.
Google’s old guidelines answered part of that question by naming categories it would not support. Its new guidelines leave more room for judgment. They point to human oversight, testing, international law, human rights, and the need to minimize unintended or harmful outcomes.
That gives the company more flexibility. It also makes the interpretation of the guidelines more important. Without the earlier explicit bans, the practical meaning of the policy will depend on how Google applies the new standards to real contracts and deployments.
The move also shows how fast the AI defense market is changing. Google is now positioned closer to other AI companies that are working with defense contractors or providing AI systems to government defense and intelligence customers.
For readers following artificial intelligence, the key takeaway is simple: AI ethics policies are no longer just about research principles or consumer products. They are increasingly tied to national security, military systems, surveillance concerns, and the race for AI leadership.