Google AI Researchers Want Google to Remove Contract Clauses They Say Could Silence AI Warnings

Google DeepMind employees are pressing the company to remove contractual restrictions they say could prevent them from warning the public about artificial intelligence risks, escalating a dispute over how freely researchers can discuss the technology they build.

Some London-based employees have contracts barring them from criticizing Google even after leaving, according to an exclusive Washington Post report based on a reviewed employment contract and interviews with two current workers. The employees spoke anonymously because they feared retaliation.

Google says it does not enforce those restrictions against former employees and has changed its terms for new hires. But workers want the language removed from existing contracts, arguing that assurances from executives leave uncertainty about their ability to speak publicly.

The disagreement centers on whether researchers can explain potential harms without fearing consequences from the company developing the technology. Employees interviewed by the outlet said the restrictions discourage public discussion at a moment when AI capabilities and applications are changing rapidly.

"This technology is just really consequential and it's also changing really fast," one employee told the newspaper. The dispute gained momentum after former DeepMind research engineer Sofia Liguori spoke to the BBC about her opposition to Google's military AI work. Her decision to speak despite contractual restrictions prompted current employees in Britain to examine their own agreements.

Some discovered similar non-disparagement provisions, and dozens subsequently participated in an internal messaging discussion about the language, one employee said. Responses included criticism of the restrictions and an emoji referencing George Orwell's dystopian novel "1984."

For the employees seeking revisions, the concern extends beyond whether Google would ultimately pursue enforcement. One worker told the Post that even a restriction they believed unenforceable could frighten colleagues into withholding information.

Google offered a different account of its approach to employee concerns. "We've always encouraged an open dialogue about safety and responsibility," a spokesperson told the Post. "Recently, we've updated our terms for new hires, which included removing the post-termination aspect of non-disparagement clauses."

The company also said employees have internal channels through which they can raise concerns with senior leaders. A Google executive responding to the internal discussion said the company would not enforce the clause against workers who had departed.

Those statements have not satisfied the current employees interviewed by the newspaper. They want Google to remove the provision entirely from existing employment agreements, rather than limit contractual changes to incoming staff.

The controversy follows public departures by researchers concerned about the direction of AI development and its military applications. Liguori's BBC interview brought particular attention to whether employees can challenge those decisions without jeopardizing their careers.

It also adds another chapter to Google's history of employee activism over defense work. In 2018, employee opposition helped persuade Google to stop pursuing a Pentagon contract involving software used to analyze drone imagery. The company subsequently introduced restrictions on certain military applications of AI, which it later dropped.

In 2024, Google fired employees who participated in office protests against its work with the Israeli government and military, the Post reported. The current dispute concerns employees' ability to discuss the consequences of that work publicly, including after their employment ends.

Workers seeking contract changes argue that their technical knowledge gives them a responsibility to contribute to decisions about AI's future. "The people who work on this technology have a role to play in shaping where it's going," one employee told the Post. "It's not just another technology. It has potentially very large impacts on humanity and democracy."