May 11, 2026
What Google DeepMind’s Union Vote Tells Us About Leadership in the Age of AI
Google DeepMind’s union vote was not about pay or working conditions. Last week, 98% of employees at the company’s London office voted in favour of unionisation over the ethics of what their AI is being used for — and over an organisation that has systematically removed every other mechanism employees had to voice their concerns.
If recognised, it would be the first unionised workforce at any frontier AI lab in the world.
It would be far too easy to dismiss this as yet another AI ethics story and just move on. But we need to look deeper, because this is a story about what happens when human-centred leadership fails completely, and at scale.
Two Kinds of Employers
Herzberg introduced us to the Two Factor Theory on employee motivation. Later, Douglas McGregor’s Theory X and Theory Y argued that there are two contrasting sets of assumptions managers hold about employee motivation.
Theory X managers assume that people are inherently lazy, dislike work and require close supervision. Theory Y managers, on the other hand, assume people are self-motivated and generally enjoy work and, therefore, seek responsibility. I would take these theories a step further and posit that there are only two types of employer when it comes to employee voice.
The first type genuinely cares. They listen, they act, and when they get it wrong, they say so. Employee voice is a live input into decision-making and not a tick-box exercise or managed output.
The second type wants to be seen to care. They build the structures – they invest in the technology, host the listening sessions, establish the ethics committees and publish principles and guidelines. But those structures exist to merely demonstrate care instead of exercising it. When employee voice conflicts with commercial interest, commercial interest wins, every single time.
Google DeepMind, it appears, falls into the latter category and the evidence has been accumulating for eight years.
Anatomy of a Carewashing Failure
In 2018, thousands of Google employees protested Project Maven, a Pentagon contract using Google AI to analyse drone surveillance footage. Google walked away from the contract, published AI principles pledging not to develop weapons or surveillance tools, and built a dedicated ethics team. It looked like proof that employee voice could shape the moral boundaries of a tech giant.
Later, the dismantling began.
The co-leads of that ethics team were fired in 2020 and 2021, respectively. Google’s published AI principles – the very document that employees had used as internal grounds to challenge military projects – were edited in February 2025 to remove the weapons pledge. Last week, despite more than 600 employees – including directors and VPs – signing a letter directly to the CEO, Google signed a classified Pentagon AI deal for ‘any lawful governmental purpose.’
The Harvard Business Review defines ‘carewashing’ as the misalignment between what leaders portray as culture of care and employee’s actual lived experience. Even though this term was initially developed in relation to wellness programmes, it applies equally to digital transformation, AI ethics and overall company culture.
What has happened at DeepMind is clearly carewashing and the cost has been compounding since 2018.
The AI Cultural Debt is Now Due
Deloitte coined the term ‘AI cultural debt’ to describe what accumulates when organisations deploy AI without addressing the human cost. A deferred reckoning on trust, psychological safety, and employee agency that grows quietly in the background, exactly like technical debt in software.
Google has been running up its AI cultural debt for near a decade now. Every fired ethics leader, every removed ethics principle, every ignored letter, all added to the balance. Last week, the invoice arrived in the form of a vote to unionise.
Note that a union was not the employees’ first choice. It was their last resort. When you remove the ethics team, you remove one voice mechanism. When you remove the published ethics principles, you remove another. Then, when you ignore 600 signatories, you remove yet another. Collective bargaining is what remains when everything else has been taken away.
Human-centred Leadership in the Age of AI: The TRUST Protocol
It’s become imperative that we develop a clear framework to guide business leaders on what human-centred leadership needs to look like in the age of AI.
For the past year, I have been working on such a framework. It’s informed by neuroscience, organisational psychology and hard leadership lessons from the organisations I have worked with over 25 years. I call it the TRUST Protocol.
- Transparency: in what AI is being used for, why, how it benefits the organisation, and when AI is used in decisions affecting people.
- Reciprocity: ensuring that AI augments human capability instead of merely extracting or replacing it.
- Uniqueness: actively protecting what humans do that AI cannot – such as empathy, judgement, moral reasoning, creativity and innovation.
- Sovereignty: defending human agency in both the design and completion of work.
- Tending: treating company culture as a living ecosystem to be genuinely cared for, not a machine to be optimised.
Every one of these principles represents an inflexion point where Googles leadership made the wrong choice.
They did not disclose the role their AI would be playing in military decisions and government contracts or the ethical implications thereof. They replaced human ethical judgment and objections with commercial calculations. And they dismantled the very structures that protect moral reasoning at work. Then they removed employee agency, step by step, and stopped tending to their culture the moment that tending became commercially inconvenient.
Meeting The Moment
The DeepMind story feels extreme, but the dynamics it exposes, are not. Every organisation deploying AI right now is making a daily choice about whether employee voice is a genuine input or a managed output. They are deciding whether ethics structures exist to guide decisions or to defend them and whether culture and care is something they actively tend and build or merely something they perform.
The difference between the two types of employers isn’t values, it’s courage. Caring genuinely and operationally, even when it’s commercially inconvenient, takes courage. Words come easy, but your employees are watching your actions and, unlike Google’s, they may not wait 8 years to act.
