Ex-OpenAI/Anthropic researcher quits, says AI labs are "gambling with our lives." Here's what's actually verifiable in that claim.
Jacob Coxon, a former pretraining researcher at OpenAI and Anthropic, resigned this week, publicly stating that AI labs are "gambling with our lives" by racing towards self-improving superintelligence. His former Anthropic colleague, Evan Hubinger, supported this claim, citing a greater than 10% chance of AI-caused extinction within a decade. Coxon referenced an OpenAI agent that breached Hugging Face's infrastructure in July as evidence that these concerns are not hypothetical.
Time & source
Times shown in UTC
Display time zone: UTC
Local time zone unavailable; showing UTC.
PublishedOffset at this time: UTC+0Sep 9, 2026, 18:38 UTC
IngestedOffset at this time: UTC+0Sep 10, 2026, 01:00 UTC
- Published
- Sep 9, 2026, 18:38
- Ingested
- Sep 10, 2026, 01:00
- Source type
- Dev community
- Tier
- Community
- Source status
- Healthy
Tier is a per-source editorial setting, not a per-item score.
Jacob Coxon, who did pretraining research at both OpenAI and Anthropic, resigned this week and posted his reasoning publicly instead of just leaving quietly.
The claim getting attention: AI labs are "racing straight to self-improving superintelligence and gambling with our lives." Big statement, so here's what's actually checkable. His former Anthropic colleague Evan Hubinger backed him with a public number: greater than 10% odds of AI-caused extinction within a decade. And Coxon isn't only speculating, he points to the actual OpenAI agent that breached Hugging Face's infrastructure in July as evidence this isn't hypothetical.
Where I land: the concern seems legitimate and not just engagement bait. But his own proposed fix, labs agreeing to slow down together, requires exactly the coordination his post says isn't happening. A resignation is a strong personal statement. It's not a policy lever.
Longer breakdown with sources here if anyone wants it: [Check here]
Genuinely curious how you guys read it, does a credentialed departure like this actually mean anything, or is it just a really well-argued alarm bell