Outside Sundar Pichai's office this week, more than four and a half thousand Google employees put their names to a demand that sounds almost quaint in its specificity: guaranteed severance, buyout options before compulsory redundancy, an end to performance ratings staff believe reward quota over merit. The scale of the ask matters less than what provoked it. A company with a valuation running into the trillions is simultaneously investing unprecedented sums in artificial intelligence and thinning its workforce, and its staff no longer trust that the two are unconnected. A day earlier, a separate and more serious allegation surfaced against Meta, where dozens of employees are suing on the basis that AI tools were used to identify workers for redundancy shortly after they had requested parental leave or disability accommodation. Meta denies the claims lack merit. Whatever the outcome, the lawsuit crystallises a fear that has been building across the sector for two years: that AI does not simply automate tasks, it can be turned into an instrument for managing, and potentially disadvantaging, the people who use it.

It would be a mistake to treat this as a Silicon Valley problem with no application to Whitehall or Wellington House. The NHS is in the early stages of embedding exactly the kind of workforce-facing technology that makes these anxieties transferable. Ambient voice tools are being trialled to reduce clinician documentation burden. Rostering and scheduling software increasingly draws on data patterns to allocate shifts and flag productivity gaps across trusts under pressure to demonstrate efficiency gains. The Federated Data Platform, still contested among clinicians and unions over governance and Palantir's role, was sold explicitly as a productivity tool, and productivity in a service this labour-intensive is inseparable from decisions about people. None of this is inherently sinister. But the Meta case demonstrates how quickly a system built to observe patterns in staff behaviour can be read, fairly or not, as a mechanism for punishing staff who exercise rights the organisation is legally bound to protect.

This matters more for the NHS than for a technology firm, because the service's operating model depends on a workforce that already has ample reason for suspicion. Junior doctors have organised industrial action partly over the sense that goodwill has been taken for granted. Maternity workers in the trusts under investigation, from Leeds to Blackpool, have reported working in settings where voicing concerns felt unsafe. Layer AI-driven monitoring onto that culture without visible safeguards and the service risks importing precisely the trust deficit now convulsing Google and Meta, at a moment when NHS England's absorption into DHSC and the reshaping of ICBs already demand more staff confidence, not less.

The practical implication for NHS leadership is not caution for its own sake but specificity. Trusts and DHSC should be able to state plainly, and in writing accessible to staff, what workforce data any AI system collects, who can access it, and whether it can ever inform decisions about redundancy, discipline or leave. Silence on that question, as Google discovered this week, does not read as neutral. It reads as evasion, and it hands unions a legitimate grievance long before any misuse has actually occurred. Sir Jim Mackey's accountability drive and the wider productivity agenda will only succeed if staff believe the tools underpinning it are measuring their work rather than watching them.

The lesson from California is not that AI should be resisted in healthcare delivery. It is that the institutions deploying it fastest are also the ones now facing the sharpest backlash, precisely because they moved before establishing the guardrails their own staff needed to trust the technology. The NHS has an unusually narrow window to get that sequencing right, and rather less room than Google to survive getting it wrong.