UK‑based Google DeepMind staff call for ending non‑disparagement bans on post‑exit warnings about AI risks
deepmind google
| Source: Techmeme | Original article
UK-based Google DeepMind employees are urging company leaders to eliminate non‑disparagement clauses that prevent them from publicly warning about AI risks, even after they leave.
DeepMind staff in the United Kingdom have pressed senior management to scrap a “non‑disparagement” clause that bars employees – even after they leave the company – from publicly warning about AI risks. According to sources cited by the Washington Post, the clause, which appears in new employment contracts, would prevent researchers from speaking out about safety concerns, ethical dilemmas or potential misuse of DeepMind’s technology.
The demand comes amid a broader push by DeepMind workers to unionise and to gain contractual veto power over military applications of their models. Employees have highlighted a recent Pentagon agreement that would allow Gemini‑family AI systems to run on classified defence networks, a move they say contravenes Google’s longstanding AI‑ethics commitments. Union‑organising efforts, backed by the Communications Workers Union and Unite, now involve roughly 300 staff members seeking formal recognition.
Why it matters is twofold. First, the clause could silence internal expertise at a time when the industry faces mounting scrutiny over safety, bias and geopolitical implications of powerful models. Second, the dispute spotlights a growing tension between large AI labs’ commercial and defence contracts and the conscience of their research teams, echoing similar labour‑rights battles in the tech sector.
Looking ahead, DeepMind’s leadership is expected to respond to the union’s demands and to the call for contract revisions. The outcome could set a precedent for how AI firms handle employee speech rights and ethical oversight, especially as governments worldwide tighten AI safety regulations. Observers will watch for any formal amendment to the contracts, the result of the union recognition vote, and whether the Pentagon partnership proceeds unchanged. The episode may also influence other AI labs grappling with “conscience clauses” and the broader debate over corporate responsibility in advanced AI development.
Sources
Back to AIPULSEN