Google DeepMind staff push back on lifetime non-disparagement clauses

Google says it changed the wording

The concern is not limited to two people. According to one employee, dozens of colleagues recently joined a discussion about the contracts on an internal communication platform. Some openly criticised the clause, while others posted emoji referring to "1984".

Google says it does not enforce such restrictions against former employees. The company also says that in its latest update to the terms it removed, for new hires, the wording that extended the ban on criticism beyond the end of employment.

That is exactly where the conflict lies. Current staff want the same change applied to earlier contracts. A verbal or public assurance that a clause will not be used is not the same as deleting it from a signed agreement. Google adds that it encourages open discussion of safety and responsibility and provides internal channels for raising concerns.

Why it matters

The story has two sides. An employee who criticises an employer is not automatically a neutral arbiter: they may misjudge risks, lack the full business context or simply disagree with a decision that is a legitimate corporate choice. Neither silencing staff nor turning every internal disagreement into a public leak is a sensible answer.

Still, the current model of AI development asks society for a great deal of trust. Companies build the models, run much of the testing, set some of the rules for releasing them and hold the most information about how their systems behave. The irony is hard to miss: Google is trying to convince the world that its AI can be trusted, while some of the people building it are fighting for a formal guarantee that they can tell the public when it should not be.