Google DeepMind employees are challenging lifetime non-disparagement clauses they say could discourage them from warning the public about AI risks after leaving the company, according to a November 3, 2026, report by The Washington Post.
The Post reported that some UK employment agreements contain restrictions on criticizing the company that extend beyond employment. Workers quoted in the report want those provisions removed or clarified in their existing contracts.
Google says it does not enforce the post-employment restrictions and has removed that aspect of its terms for new hires. That response addresses part of the dispute, but leaves an important distinction: a company’s stated enforcement policy and the language in an employee’s signed agreement are different things. The employees’ request concerns the latter, too.
What The Post Reported About the Contracts
The Post’s reporting drew on a review of contracts and interviews with current employees. According to NDTV’s account of the investigation, two employees spoke anonymously because they feared action from the company.
The reported provisions are non-disparagement clauses. The concern is that criticism of Google could be restricted for the rest of an employee’s life, including after departure. Employees told the Post that this language could discourage public discussion of the risks of AI systems they helped develop.
The full contract wording has not been independently reviewed for this article, and the cited reporting provides no legal determination about enforceability. The evidence concerns particular UK agreements, not a verified description of every DeepMind employee’s contract worldwide.
The distinction between criticism and a safety warning is central to the dispute. A researcher explaining a risk in a company’s AI work could also be saying something damaging to that company’s reputation. That potential overlap helps explain the employees’ concern, without assuming that every public warning would fall within the reported clause.
The workers’ argument is therefore about the effect of having the restriction in writing, as well as any possibility of enforcement. According to the reporting, they worry that the wording could influence whether a current or former employee decides to speak publicly in the first place.
A Former Employee’s Military AI Criticism Prompted Discussion
Former DeepMind employee Sofia Liguori’s public opposition to Google’s military AI work brought the issue to wider internal attention, the Post reported.
According to NDTV’s summary, Liguori told the BBC in October that she had ignored restrictions in her contract when speaking out. Current UK employees subsequently checked their own agreements and found similar lifetime non-disparagement provisions.
Dozens of employees then discussed the language on an internal messaging platform, according to the report. Some criticized the clause directly; others responded with an emoji displaying “1984,” a reference to George Orwell’s novel.
That is a documented internal reaction, not evidence of a company-wide consensus. The reporting identifies employees participating in a discussion, but does not establish what proportion of DeepMind’s workforce shares their position.
The sequence matters because it connects a broad concern about AI transparency to a specific employment document. Liguori’s reported decision to speak publicly prompted colleagues to examine what their own contracts said. Their resulting demand was concrete: address the provisions in agreements people had already signed.
Liguori’s account and the internal response are reported through the Post and subsequent coverage. They are not based on an interview conducted for this article.
Google’s Response Covers Two Different Issues
Google’s response combines an assurance about enforcement with a change to the terms offered to new employees.
The company said it does not enforce the post-employment restriction at issue. NDTV also reported that a Google executive conveyed that position in the internal discussion.
A spokesperson described the contract update this way, as quoted by NDTV:
“Recently, we have updated our terms for new hires, which included removing the post-termination aspect of non-disparagement clauses.”
The scope of that statement is important. It names new hires and the post-termination aspect of the clauses. It does not announce the removal of every non-disparagement provision or an amendment to every existing agreement.
| Issue | Google’s reported position | What the response leaves open |
|---|---|---|
| Enforcement after employment | Google says it does not enforce the restriction. | Existing employees still want clarity about the text in their agreements. |
| Terms for new hires | The post-termination aspect has been removed. | The statement does not say all non-disparagement language has been eliminated. |
| Existing contracts | Employees are seeking removal or clarification. | The cited reporting does not describe a blanket revision of those agreements. |
This is the most consequential detail in the response. New employees receive revised terms, while employees with the older wording are being given an assurance about how Google says it handles that wording. Those are different forms of reassurance.
My view is that the new-hire change is meaningful, but its scope should stay precise. It removes the post-employment component from the terms Google says it now offers new starters. Describing that as a complete resolution would overlook the employees asking what happens to contracts already in place.
The statement also should not be read as a general policy covering every possible employment restriction. It was given in response to this non-disparagement dispute.
Why Existing Wording Still Matters to Employees
Employees quoted in the reporting want Google’s position reflected more clearly in their existing agreements. Their concern is that language can discourage speech even when a company says it will not act on it.
Frequently Asked Questions
3 questions
1Which DeepMind Contracts Are Involved?
The Post reported on some UK Google DeepMind employment agreements. The available reporting does not establish that every DeepMind employee has the same terms.
