DeepMind researchers fight lifetime non-disparagement terms to speak about AI risk

Staff at Google DeepMind are pushing back on employment terms that include lifetime non-disparagement clauses. Those clauses would prevent current and former employees from publicly criticizing Google indefinitely. The researchers argue that this effectively silences them on AI safety concerns, which are exactly the issues the public most needs insiders to be able to discuss.

The dispute echoes the 2024 controversy at OpenAI, where departing employees faced equity clawback threats tied to non-disparagement agreements. That episode led OpenAI to release staff from the clauses and fueled calls for a 'right to warn' for AI lab employees. Several jurisdictions have since debated whistleblower protections specific to frontier AI work.

The timing is pointed. This week METR warned that Claude capabilities are improving roughly twice as fast as a year ago, largely through AI-assisted R&D. That is the kind of development where insider perspective is valuable. Labs argue confidentiality protects trade secrets, while critics say broad non-disparagement terms go well beyond protecting IP.

What to watch: whether Google narrows or drops the clauses, whether employees go public with specific concerns, and whether lawmakers cite the dispute in whistleblower legislation. Google has not publicly detailed its response in the available coverage.