Over 580 Google employees have signed a petition urging the CEO to refuse taking on classified AI work for the U.S. military.

```

A new wave of employee protests has erupted within Google. Hundreds of AI researchers have signed an open letter demanding CEO Sundar Pichai refuse to use the company’s AI systems for classified tasks commissioned by the US Department of Defense.

According to Bloomberg, organizers claim to have collected more than 580 signatures, including over 20 directors, senior directors, and vice presidents, as well as multiple senior staff members from Google’s AI research subsidiary DeepMind. Organizers stated the letter would be delivered to Pichai on Monday.

The open letter explicitly calls for Google to refuse all classified workloads, arguing that once AI tools are deployed on classified systems physically isolated from the public internet, the company will be unable to monitor their actual usage.

The letter is released as cooperation negotiations between Google and the Pentagon continue. Reports say the parties are discussing the use of Google’s AI tools for “all lawful purposes.” Critics believe this wording could, in practice, cover fully autonomous weapons systems and large-scale domestic surveillance.

In terms of the market, this employee action adds further uncertainty to Google’s future pace of defense business expansion and the boundaries of compliance.

Employee Concerns: Vague Contract Boundaries, High Risk of Losing Control over Classified Systems

The crux of the open letter is employees believe Google has failed to set clear and enforceable red lines for AI usage on classified networks.

Sofia Liguori, AI research engineer at Google DeepMind’s UK branch, said she signed the letter because Google failed to discuss AI’s use in classified areas with employees. She pointed out that once AI tools are deployed on classified systems, the company will have no technical way to track and limit how these tools are actually used.

“The company’s response has always been to ask employees to trust leadership to sign good contracts,” Liguori said. “But these statements are extremely vague.” She specifically flagged the risks of “Agentic AI”:

“The level of autonomy it can achieve is worrying. It’s like handing over an extremely powerful tool while giving up any control over how it’s used.”

The open letter reads: “Currently, the only way to ensure Google does not become complicit in such harm is to refuse any classified workload. Otherwise, such uses may happen without our knowledge or power to stop them.”

History Repeats: From Maven to Today’s Boundary Disputes

This employee action is not the first. In 2018, Google employees protested strongly against the company’s involvement in the Pentagon's “Project Maven,” which aimed to use AI to analyze objects in drone video footage.

Faced with employee opposition and resignations, Google eventually established new AI principles and decided not to renew the Maven contract.

However, Google subsequently gradually rebuilt its relationship with the defense industry, and last year removed passages from its AI principles that promised to avoid using the technology for weapons or other potentially harmful applications.

The organizers of this letter stated in their announcement: “Maven is not over.” They said, “Workers will continue organizing around the issue of weaponization of Google AI until the company draws clear and enforceable boundaries.”

Google and the Pentagon: Cooperation Accelerates, Boundaries in Doubt

Meanwhile, Google’s cooperation with the Pentagon has substantially advanced. This March, Google opened its Gemini AI agents to more than three million Pentagon employees for non-classified use; previously, the Gemini chatbot had been opened in December of last year.

Emil Michael, Deputy Secretary of Defense for Research and Engineering, told Bloomberg in March that the Pentagon would start with non-classified levels of Gemini agent use, “then we will go into classified and top-secret levels,” revealing that negotiations over the use of Google AI agents in classified clouds were underway.

Google and the Pentagon are negotiating authorization for AI tools for “all lawful purposes.” This wording is criticized as possibly including weapon systems. In response, the Pentagon strongly refuted this characterization, insisting that business companies should not have veto power over usage policy in times of war or preparation for war.

This controversy contrasts with Anthropic’s recent situation—reports say the Pentagon is seeking to exclude Anthropic and its Claude AI tools from the US defense supply chain, and is actively seeking new tech giant partners, making Google a potential replacement option.

Risk warning and disclaimerThe market has risks; investments should be made cautiously. This article does not constitute personal investment advice and does not take into account specific investment goals, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article suit their particular circumstances. If you invest based on this article, the responsibility is yours. ```