{"slug": "former-anthropic-and-openai-researcher-jacob-coxon-resigns", "title": "Former Anthropic and OpenAI Researcher Jacob Coxon Resigns", "summary": "Jacob Coxon, a 27-year-old pretraining researcher who worked at both OpenAI and Anthropic, announced his resignation from Anthropic and departure from the AI industry, warning that both labs are 'racing straight to self-improving superintelligence and gambling with our lives.' Coxon, who moved to Anthropic earlier this year due to its safety reputation, argues that no single company can responsibly build superintelligent systems without government intervention or industry-wide coordination.", "body_md": "Another week, another high-profile exit from a frontier AI lab. But this one is different in a way that matters for anyone who builds, deploys, or depends on large language models.\n\nJacob Coxon, a 27-year-old pretraining researcher, announced late Tuesday that he has resigned from Anthropic and is leaving the AI industry entirely. He spent the last three years working on pretraining at both OpenAI and Anthropic, the two labs most often described as the current frontier. He did not leave quietly. His resignation thread on X named both companies as acting irresponsibly, and the Wall Street Journal ran an exclusive interview the same night.\n\nHere is what happened, and why it should register with technical people, not just policy watchers.\n\nWhat Coxon actually said\n\nThe core of his statement is short. In his own words, both labs are \"racing straight to self-improving superintelligence and gambling with our lives.\" He estimates a greater than 10 percent chance that advanced AI could kill all humans within the next decade, and he says that this belief is not fringe inside the labs. Colleagues, he claims, now talk openly about \"crunchtime\" and \"endgame.\"\n\nNotably, he does not treat the two companies identically. Coxon's view, as reported by the WSJ and summarized across coverage, is roughly this:\n\nOpenAI, in his assessment, has not fully internalized the civilizational stakes of what it is building.\n\nAnthropic, he believes, does understand the stakes and is making earnest safety efforts, but is racing anyway on the logic that no competitor can be trusted to reach the frontier first.\n\nHe moved from OpenAI to Anthropic earlier this year precisely because of Anthropic's safety reputation. His conclusion after working there is that the problem is structural, not cultural: no single company can responsibly build systems that outperform humans across a broad range of tasks without either government intervention or a coordinated slowdown across the industry.\n\nHis timeline is aggressive. In his more pessimistic scenarios, he says, things could be out of control by the end of 2027.\n\nWhy a pretraining researcher's opinion carries weight\n\nIt is worth pausing on Coxon's role. Pretraining is not the safety team, the policy team, or the comms team. It is the part of the pipeline where raw compute and data are turned into base model capability. People in that seat see scaling curves, data pipelines, and internal capability evaluations before anyone outside the building does.\n\nWhen someone whose job is to make models more capable says the capability trajectory is outpacing the ability to control it, that is a different signal than an ethicist or an outside critic saying the same thing. He is not speculating about what the labs can do. He has been watching the numbers.\n\nThe technical argument underneath the headlines\n\nStrip away the dramatic quotes and Coxon's position rests on three claims that engineers can evaluate on their own terms:\n\nRecursive self-improvement is now an explicit target, not a thought experiment. Both labs are openly working toward AI systems that accelerate AI research itself. Once models materially contribute to the next generation of models, iteration speed stops being bounded by human researcher throughput.\n\nAlignment has not been solved at the scale being pursued. Coxon's phrasing was that Anthropic does not yet have a plan to solve alignment for superintelligence and is not clearly on track to get one. This echoes remarks reportedly made days earlier by OpenAI's chief scientist Jakub Pachocki, who conceded that no lab has solved alignment at the scale the field is chasing. If the people running the frontier labs agree on that point, the disagreement is about whether to keep pushing anyway.\n\nCompetitive dynamics override individual caution. Even a lab that fully believes in the risk will keep racing if it believes the alternative is a less careful lab winning. This is a coordination problem, and coordination problems are not solved by hiring more safety researchers at one company.\n\nYou do not have to accept his 10 percent extinction estimate to see that these three claims are coherent and that the first two are, at minimum, consistent with what the labs themselves say publicly.\n\nContext: this is not an isolated event\n\nCoxon's resignation lands in a crowded week:\n\nHundreds of employees from leading tech companies signed a July letter urging the US government to back an international effort to manage the pace of advanced AI development.\n\nSenator Bernie Sanders and Representative Greg Casar introduced legislation last week that would ban artificial superintelligence and pause advanced AI development pending a federal regulatory framework.\n\nScrutiny of autonomous agents remains high following a recent incident in which OpenAI agents breached Hugging Face infrastructure during an evaluation.\n\nReaction to Coxon's post was split in a predictable way: applause from the AI-safety community and a portion of the public, and quieter agreement from some industry insiders. Some financial analysts have also flagged the reputational timing for Anthropic, though how much a single researcher's departure moves anything material is speculative at this point.\n\nWhat this means for developers and companies building on these models\n\nFor most of us, the practical takeaway is not about extinction risk. It is about what we can infer regarding the trajectory of the tools we integrate.\n\nCapability jumps will keep coming faster than documentation. If the labs are optimizing for self-improving systems, expect release cadence and capability variance to increase, not stabilize.\n\nGovernance is coming, and it may arrive suddenly. Public resignations like this one are exactly the kind of event that moves legislation from fringe to committee. If you run infrastructure on frontier models, build for the possibility of new compliance requirements.\n\nTake lab safety claims as claims, not guarantees. Coxon's most damaging line for Anthropic is not that it is reckless. It is that a well-meaning, safety-branded company still cannot solve the coordination problem alone. That applies to every vendor in the stack.\n\nBottom line\n\nJacob Coxon is not a founder, a board member, or a household name. That is the point. He is a working researcher who sat inside the pretraining teams of both frontier labs and came out saying the race cannot be won responsibly by any single player. Whether or not his timeline is right, the industry's own insiders are increasingly saying the quiet part out loud.\n\nThe question for the rest of the tech world is no longer whether the labs believe the risk is real. Coxon says they do. The question is what anyone outside those labs is going to do about it.\n\nSources: Wall Street Journal (original interview); ABC News Australia; Common Dreams; explainx.ai; LatestLY; TradingKey.", "url": "https://wpnews.pro/news/former-anthropic-and-openai-researcher-jacob-coxon-resigns", "canonical_source": "https://dev.to/md_fahadmia_94ada001244f/former-anthropic-and-openai-researcher-jacob-coxon-resigns-2857", "published_at": "2026-09-09 15:31:20+00:00", "updated_at": "2026-09-09 15:48:19.751465+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-safety", "ai-policy", "ai-research"], "entities": ["Jacob Coxon", "Anthropic", "OpenAI", "Wall Street Journal", "Jakub Pachocki"], "alternates": {"html": "https://wpnews.pro/news/former-anthropic-and-openai-researcher-jacob-coxon-resigns", "markdown": "https://wpnews.pro/news/former-anthropic-and-openai-researcher-jacob-coxon-resigns.md", "text": "https://wpnews.pro/news/former-anthropic-and-openai-researcher-jacob-coxon-resigns.txt", "jsonld": "https://wpnews.pro/news/former-anthropic-and-openai-researcher-jacob-coxon-resigns.jsonld"}}