Google says it will not enforce lifetime non-disparagement clauses against people who have left Google DeepMind. It also says it has removed the post-employment part of the clause from the terms used for recent new hires. The move followed a Washington Post report on November 3, 2026 and an internal staff discussion in London. Current employees say that is not enough, because the clause is still in their own contracts.
This post covers what was reported, what Google has and has not changed, and why a gag clause matters when the people bound by it build frontier AI. It follows our earlier report on DeepMind staff who resigned over Google's classified Pentagon AI contract, where the lifetime clause first appeared as an employee claim.
TL;DR: what is confirmed and what is open #
| Question | Answer |
|---|---|
| What changed? | Google says it does not enforce the clause against former staff. New-hire terms drop the part that applies after employment ends. |
| Who said so? | A Google spokesperson, and a Google executive in an internal discussion, as reported by the Washington Post and outlets citing it. |
| What has not changed? | Current staff say the clause remains in existing contracts. The reports do not say Google agreed to remove it. |
| Who is affected? | Some London-based Google DeepMind staff. The number affected is not stated. |
| Is the contract public? | No. The Post says it reviewed one contract. Google has not published the wording. |
Agents that act on your behalf need rules. See how explainx helps secure your agents →
What the Washington Post reported #
The Washington Post reported on November 3 that employment contracts for some London-based Google DeepMind staff include a non-disparagement agreement that applies for life, including after they leave. Two current DeepMind workers in London spoke to the paper anonymously, and the Post reviewed one such contract.
The workers said the clauses stop them from speaking publicly about the possible impacts of AI. Coverage in the New Zealand Herald says dozens of employees joined an internal discussion. Some criticised the clause directly, and some posted a "1984" emoji.
How the issue started #
In October, former DeepMind research engineer Sofia Liguori told the BBC that she was speaking out against Google building military AI. She said that, in principle, her contract did not let her criticise the company for the rest of her life. We covered that interview in our BBC-based report on the resignations.
According to the later reports, her comments pushed current UK staff to check their own contracts. Some found they were bound by the same clause. That is what turned a single employee account into an internal debate.
What Google says it changed #
Google's statement, as reported, has three parts:
- It has "always encouraged an open dialogue about safety and responsibility."
- It has updated the contract terms for new hires and removed the portion that applies non-disparagement after the employment relationship ends.
- It does not enforce these clauses against people who have left, and it has internal channels for staff to escalate concerns to senior leaders.
The Seoul Economic Daily reports the same points. It also says current employees want the clause removed from existing contracts, not only for future hires.
Note the gap. "We do not enforce it" is a promise about behaviour. It is not a change to the signed text. A clause that stays in a contract can still shape what a person feels safe saying, even if the employer never sues.
What the law may say #
The reports give little legal detail. The Post says British law protects whistleblowers who speak out in the public interest. One employee doubted that a lifetime restriction can be enforced, but said it can still push workers to censor themselves. As the Seoul Economic Daily quotes one employee, a provision that cannot be enforced legally "effectively exists to intimidate people."
We are not lawyers, and this is not legal advice. The practical point is simple. A clause can be weak in court and still work on a person who has a mortgage and a career.
Why this matters for AI #
Frontier labs hold information the public cannot get anywhere else: how models behave in testing, which safety checks were skipped, and what customers such as defence agencies are allowed to do with them. Former staff are often the only people who can say so. When a contract limits that speech for life, the public loses an early warning channel.
This is not only a Google issue. Other labs have faced the same question. Our report on OpenAI employees who warned about security lapses for years shows how much of the public record in this field depends on insiders speaking up. The same pattern runs through the debate over AI governance and safety standards and the FTC probe of OpenAI and Anthropic over safety.
The timing also matters. In the Post's account, the debate sits next to Google's work with the military. The Post notes that in 2018 Google dropped a Pentagon drone-image analysis contract after employees petitioned, and that in 2024 it fired workers who protested its work with the Israeli military. Staff who object to that work are the ones most likely to need the freedom to say so.
Non-disparagement, NDA, non-compete: three different tools #
These three terms get mixed up, so it helps to separate them.
- A confidentiality agreement (NDA) protects secrets such as unreleased models, code or customer data. It is normal at every lab.
- A non-compete limits where you can work next. In April 2025, Business Insider reported that some Google DeepMind staff in the UK were asked to sign non-competes of up to 12 months, and TechCrunch relayed the report. Google told reporters it uses non-competes "selectively."
- A non-disparagement clause limits what you can say about the company. It covers opinions and criticism, not only secrets. That is why a lifetime version worries safety researchers: it can reach a plain statement such as "I think this contract was a mistake."
The 2025 non-compete story was about talent moving between rivals. The 2026 clause story is about speech. Different problem, same employer, and both show how much legal structure sits around a frontier lab workforce.
What the Google case shows about how these changes happen #
The sequence in the reports is short. A former employee spoke to the BBC. Current staff checked their contracts. An internal thread grew. A national newspaper asked questions. Within days, a Google executive told staff the clause would not be enforced against leavers, and Google described the new-hire change.
That pattern is useful for readers who work at other labs. A written change to a contract is slower than a verbal assurance, and the verbal assurance can come quickly once outside attention arrives. Whether the written change follows is the part to watch.
What current staff want #
The employees quoted ask for one thing: remove the clause from existing contracts, not only new ones. As of the reports on November 3 and 4, the sources do not say Google agreed. They do not give a deadline, and they mention no formal letter or petition. The discussion took place on an internal messaging platform.
What to watch next #
- Whether Google changes the signed contracts of current staff, not only new-hire terms.
- Whether Google publishes the clause wording or a written policy.
- Whether other labs with similar contracts face the same questions from staff or regulators.
- Whether UK lawmakers or regulators comment on lifetime non-disparagement in tech contracts.
What to do if you work in AI #
If your contract has a non-disparagement or confidentiality clause, read the exact wording first. Check whether it has an end date. Check whether it carves out protected disclosures to regulators or in the public interest. Ask for the company's reading in writing, and talk to an employment lawyer or a whistleblower support group in your country before you speak in public.
For teams running AI agents, the same logic applies to systems: you want early warning when something goes wrong. [AgentBeam](https://agentbeam.com/?ref=explainx) is the agent security platform from the explainx.ai team that stops AI agents before they take dangerous actions.
Frequently asked questions #
Is Google still enforcing the clause? Google says it does not enforce it against people who left. It says new-hire terms no longer include the post-employment part. Current staff say the clause is still in their contracts.
Does this change anything for the public? Possibly. If former staff feel safe to talk, more first-hand accounts of lab practice can reach the public. That depends on how far people trust a verbal assurance.
Where can I read more on this story? Start with the Washington Post report and our earlier post on the BBC reporting.
Related reading #
Details reflect press reports as of November 4, 2026. Contract wording has not been published, and this is not legal advice.