A clause with no leaving date
The wording the Washington Post describes at Google DeepMind in London has one unusual feature: it sits in the employment contract itself and has no end date. Some staff, the Post reports, are bound by non-disparagement obligations that continue after they leave the company, potentially for the rest of their lives. The account rests on a contract the Post reviewed and interviews with two current DeepMind workers, who spoke anonymously to avoid retribution.
The workers' objection is specific, and it is worth keeping specific. They told the Post the clause blocks them from communicating with the public about the potential impacts of AI. This is not a dispute about office grievances or pay. The people building the systems are saying a non-disparagement restriction with no stated end date, signed as a condition of employment, is what stands between their knowledge and the public debate about the technology. One of them put the public-interest case plainly: the technology is consequential and changing fast, and often this is the only way people find out about the harms.
How the clause surfaced matters too. Former DeepMind employee Sofia Liguori told the BBC last month that she had chosen to ignore the restrictions in her own contract to speak against Google building military AI. Colleagues in Britain then checked their contracts, and some found the same wording in theirs. Dozens joined a thread on an internal messaging platform. Some responded with "1984" emojis, according to one of the workers. That detail is small and exact, and it tells you how the staff reading the clause understood it.
Two answers, neither of them a new contract
Google has now answered twice, and the two answers are narrower than they first sound.
On Friday, an executive responded in the internal thread and said the company would not enforce the clause for workers who had already left. On Tuesday, a spokesperson went a step further on paper, for a different group of people: "Recently, we've updated our terms for new hires, which included removing the post-termination aspect of non-disparagement clauses." The spokesperson added that the company does not enforce the clauses after people leave, and pointed to internal channels for raising concerns.
Read those two statements together and the remaining group is easy to identify: current employees who signed the earlier contracts. Former employees have been given an assurance that the clauses will not be enforced after departure. New hires receive updated wording. But existing employees have not been told that their contracts will be amended, and Google's statements do not establish what would happen if someone still employed publicly criticized the company. The staff who spoke to the Post are asking for a third step: removing the wording from existing contracts entirely.
The distinction is not pedantry. One employee told the Post they did not think the lifetime restriction was enforceable. The report also notes that British law provides protections for qualifying whistleblowers who disclose information in the public interest. But no tribunal ruling on this particular clause is reported. The employee's concern was that enforceability might not be the point: "If it's not enforceable then the reason is to basically intimidate people." A contractual restriction can influence whether someone speaks even before anyone attempts to enforce it, because the employee has to consider the possibility of a legal dispute and its consequences. A promise not to enforce lowers that price. Removing the words removes the question.
The comparison Google will not enjoy
There is a related precedent at a rival, although the contractual circumstances were different. In 2024, reporting revealed that OpenAI's departure arrangements tied some former employees' vested equity to perpetual non-disparagement obligations. Following public pressure, OpenAI released former employees from those existing obligations and removed the clauses from its departure paperwork, as CNBC reported. AI Weekly draws the comparison with the DeepMind dispute. The distinction is important: Google's current employees are asking for existing employment contracts to be changed, not simply for a future hiring policy.
AI Weekly also notes that outside researchers have started making the same demand in public. Seán Ó hÉigeartaigh wrote on X that he had already confirmed the in-perpetuity clauses with multiple people, and Nathan Calvin urged Google to release prior employees who had signed from those obligations.
Context inside Google cuts the same way. The Post notes that executives and staff at Anthropic and OpenAI comment publicly about AI risk as a matter of routine, while fewer Google voices join that debate, and that in September Demis Hassabis himself joined the chief executives of OpenAI and Anthropic in calling for a slowdown in AI development to allow more safety research. A company whose former AI chief can call publicly for a pause, while some of its researchers hold non-disparagement clauses with no end date, has a gap between its public position and its paperwork. Closing that gap requires editing contracts, not issuing statements.
What to watch
Three things would settle this. First, whether Google amends existing contracts or keeps relying on non-enforcement. Second, whether any current employee speaks publicly about AI risk and tests the assurance while still employed, the one situation neither statement quite covers. Third, whether the number bound by the clause ever becomes public, because "some staff in London" is currently the whole of the known scope. Until the first of those happens, the accurate summary is the one the staff themselves are giving: the clause is still there, and they are still asking for it to go.
Sources
- [1] Gerrit De Vynck, The Washington Post, "Google researchers push to end non-disparagement clauses to tell what they know about AI", November 3, 2026, read in full via Missouri Lawyers Media syndication, not the Post original behind its paywall (the contract reviewed by the Post and the two anonymous current workers; the non-disparagement wording continuing after leaving; the internal thread joined by dozens of staff and the "1984" emoji posts; the Liguori BBC trigger; the employee quotes on public knowledge, enforceability and intimidation; the Friday executive response on non-enforcement for leavers; the Tuesday spokesperson statement on updated new-hire terms and non-enforcement after people leave; the staff demand to drop the clause from existing contracts; the September Hassabis slowdown call and the Google protest-history context)Read source
- [2] AI Weekly, "DeepMind staff push to drop lifetime non-disparagement clauses", November 3, 2026 (read first-hand in full before publication — corroboration of the no-end-date scope, the Friday executive response and the new-hire update, the staff demand for removal from existing contracts, the public posts by Seán Ó hÉigeartaigh and Nathan Calvin, and the explicit parallel to OpenAI's 2024 release of former staff; also the honest limits: the Post reports no count of bound staff and no tribunal test)Read source
- [3] CNBC, May 2024, via Slashdot, "OpenAI Releases Former Employees From Controversial Exit Agreements" (the 2024 precedent as reported at the time: OpenAI would not cancel vested units, would remove non-disparagement clauses from departure documents, and would release former employees from existing non-disparagement obligations unless the provision was mutual)Read source