Robert O’Callahan announced his resignation from Google on September 24, 2026, citing concerns about the pace of AI development and the direction of his work.
O’Callahan, who worked on improved tools for hardware chip design at Google Deepmind, stated that while he enjoyed his role and work environment, he became uncomfortable with the primary impact of his tools: accelerating the design of AI chips that would make artificial intelligence cheaper, faster, and more pervasive.
“My team’s goal is ultimately to make AI much cheaper and lower-latency, and I don’t think that’s good for people right now,” O’Callahan wrote in his resignation email. “I firmly believe AI progress is currently far too rapid.”
O’Callahan expressed multiple concerns about AI development. He stated that the current rate of change is unsustainably high, preventing individuals and societies from adequately understanding and adapting to AI. He noted that “AI is developing faster than humans can individually and collectively understand it and adapt to it,” comparing the pace unfavorably even to previous technological shifts like the personal computer, internet, and smartphone revolutions.
On existential risks, O’Callahan acknowledged that while he is not fully convinced of certain doom scenarios, he believes the risks deserve serious consideration. He noted that phenomena predicted by AI risk researchers—including “reward hacking, misalignment, deceptive models, model eval awareness, psychotic swarms”—have already materialized.
O’Callahan also expressed concern about broader AI-related harms beyond existential risk, including “cognitive surrender, AI-induced psychosis and loneliness, power concentration, economic disruption, cybersecurity, lack of accountability.”
The timing of his departure was not driven by recent AI slowdown discussions or viral resignation trends, according to O’Callahan, but rather by a long-planned personal commitment: a ten-day backpacking trip beginning the following Monday.
Moving forward, O’Callahan said he plans to continue maintaining Pernosco and rr (his existing projects) and investigate how AI agents debug code and whether debugging tools could assist them. He stated his intention to ensure future work is “unambiguously pro-human.” O’Callahan noted that his position in New Zealand, outside both the tech industry bubble and the American sphere, has shaped his perspective on these issues.
Key facts
- O’Callahan worked on hardware chip design tools that he recognized would primarily accelerate AI chip design and development
- He cited concerns that current AI progress is too rapid for human societies to adequately understand and adapt to
- O’Callahan noted that predicted AI risk phenomena including reward hacking, misalignment, and deceptive models have already occurred
- His resignation was scheduled before recent viral discussions about AI slowdown or similar resignations
- He plans to continue maintaining his existing projects Pernosco and rr while investigating AI debugging capabilities