Researcher leaves Google DeepMind after AI ethics leaders cave under pressure
Alex Turner recounts his attempts to persuade Google and leading AI researchers to honor public commitments
I recently listened to a podcast episode that defined modern cynicism as being aware that something is wrong but still willingly participating in it. This makes systems pretty hard to change.
In a detailed account published yesterday of the reasons he left Google DeepMind, researcher Alex Turner teaches us that such cynicism is alive and well. But he also shows us how to fight it.
The piece chronicles his efforts, in the wake of human rights abuses by ICE and the Anthropic/Pentagon dispute, to get AI ethics advocates to act in accordance with the principles they espouse. Many of these people, including renowned AI scientist Stuart Russell and Google Gemini lead Jeff Dean, have been publicly outspoken on certain principled stances related to Big Tech and government. They have considerable leverage and influence. Yet, according to Turner, they caved in the face of corporate and political pressure.
In Turner’s words:
This essay tells the story of why I left Google DeepMind. It is also the story of something larger: how powerful people and institutions failed, one after another, to keep their AI ethics promises in the face of pressure.
I won’t go into all the examples; Turner explains them better than I can. But a central thread is Turner’s efforts to get Google to refuse a government deal that went against a red line they had previously drawn: that their tech would not be used for autonomous weapons. While this red line had already been rolled back, the company, and many of its senior leaders, remained signatories of a 2018 pledge to “neither participate in nor support the development, manufacture, trade, or use of lethal autonomous weapons.” Chief Scientist and lead of Google AI Jeff Dean even reaffirmed his position in a 2026 tweet.
A few things stick out to me about the piece, which I found both depressing and awe-inspiring.
Perhaps most obvious is further confirmation that we can’t trust the AI companies, even the ones who say they take safety seriously. To use Turner’s words:
When someone signs “I will not support the development of lethal autonomous weapons,” then stays while their company sells unrestricted AI to a military that wants exactly that, they teach every counterparty a lesson: these safety people will not act, even at their own brightest line. The next commitment they make is worth less. Eventually it’s worth nothing.
And indeed, we’ve seen this happen time and time again. Even Anthropic, the AI company widely perceived as the most ethical, rolled back its 2023 promise to stop developing frontier AI systems if adequate safety measures couldn’t be demonstrated, saying it no longer made sense due to competitive pressures. OpenAI began as a company committed to “saving the world” and was founded with governance mechanisms designed to keep its mission from being corrupted by profit-driven motives. These governance mechanisms were severely weakened over time, and OpenAI has gradually transitioned from a non-profit to a more conventional, investor-funded company.
But there’s also a strong lesson here about moral courage. While Turner doesn’t ultimately succeed in swaying leadership, he has some pretty impressive successes merely by acting courageously and strategically. These include putting a draft of an oversight framework for government AI contracts in front of Google DeepMind CEO Demis Hassabis, building support coalitions of employees, and getting the head of Google AI, Jeff Dean, to sign an amicus brief that made the White House think twice about the Google deal. When Turner is ignored, he follows up. When one avenue fails, he tries another.
As a reminder to himself to keep “breaking bounds,” he keeps a photo on his phone of Alex Pretti, the Minneapolis nurse who was shot by ICE in January. (Google has delisted apps that alert people to ICE activity and sells its Cloud services to DHS. Turner’s saga begins with trying to right these injustices.)
While the big players featured in his account deliberately try to abdicate ethical responsibility through justification or avoidance, Turner chases it. That’s a quality all of us could use. What would it look like if Anthropic had this quality? The company broke its promise to stop developing frontier systems in the absence of demonstrable safety mechanisms because it reasoned other companies would continue development anyway. But a company that truly cared about these commitments, that was willing to chase ethical responsibility for the greater good, would have asked: “how can I get my competitors to stop too?” It would have used its influence to try to build a coalition and to try to influence policy.
Turner’s actions in the Pentagon/Anthropic dispute set an example of such attempted coalition building. When Anthropic was dropped by the Pentagon over its refusal to agree to the “all lawful use” clause, Turner tried to figure out how to support Anthropic’s stand, knowing that Google’s acquiescence would undermine it:
The main question which weighed on me: can I stop Google from caving, from accepting an “all lawful use” deal? If Anthropic says “no” and Google also says “no,” now we’re getting somewhere. That seemed hard. I wanted to make it happen anyway. Google could cave at any moment…
I also appreciated his response to the common “at least I have a seat at the table” justification:
But should [senior employees] not stay to keep steering in a positive direction? To this I must object: “What steering?” [The “all lawful use” deal with government] may have been the clearest red line Google’s Gemini project will ever face… If their “seat at the table” couldn’t produce a single binding provision in that situation, then when would it?
It would have been so easy for Turner to write an email or two, conclude he’d done what he could, and move on. The fact that he kept going, and that his actions did actually influence extremely senior people (even if they didn’t always follow through) is a knock against cynicism in an increasingly cynical world. We need more of those.
The analyses and opinions expressed on AI StopWatch reflect the views of the individual contributors and the sources they cover, and should not be taken as official positions of the Machine Intelligence Research Institute.