The public argument over AI risk has widened beyond one researcher's departure. The National Post brings together warnings from several people working on frontier-model safety, alongside a call for international coordination over automated AI development. That makes this a report about a growing debate, not a prediction that disaster has been measured or scheduled.
Jacob Coxon Watch analysis
What happened
The newspaper reports on former Anthropic researcher Jacob Coxon's warnings about companies racing towards self-improving systems. It also describes statements by Evan Hubinger, Samuel Marks and Paul Christiano. The important distinction is between these researchers' attributed judgements and established forecasts: their concerns deserve scrutiny without being converted into a countdown clock.
Why it matters
This is more useful than treating every warning as another instalment of the same alarm. Named researchers are making different kinds of claims: about incentives, alignment progress, the speed of automated research and the institutions needed to respond. Readers should be able to examine those claims together without assuming that agreement on concern means agreement on probability, timing or policy.
The report is also an opportunity to distinguish evidence from authority. Relevant expertise matters, but it does not make a personal probability estimate independently verified. Nor does a large number of signatories settle which restrictions would work or what they would cost. Those are questions for evidence and argument, not arithmetic applied to a petition.
Our read
The worthwhile development is the breadth of the public discussion and its proposed coordination mechanism. This deserves its own place in the conversation, rather than disappearing as another reply beneath an earlier resignation story. Take the researchers' arguments seriously, keep their uncertainties attached, and ask what observable evidence would support or weaken their recommendations.
What to watch
- Whether the researchers explain the assumptions behind their risk estimates.
- How proposals to pace automated development would be implemented and checked.
- Whether companies answer the specific concerns about incentives and alignment progress.
- Which claims receive independent evidence rather than further repetition.
Discussion spark: What evidence would persuade you that frontier AI development should speed up, slow down or change direction?
Sources and evidence
- Here are all the recent warnings about how AI 'could kill us all' within a decade – National Post (10 September 2026, 17:57 UTC)
- Anthropic Threat Report Shows AI Bioweapon Research Risks (10 September 2026, 17:57 UTC)
- Former AI researcher: Not enough time for employees to consider safety of models (13 September 2026, 14:52 UTC)
Independent WittyWires-operated tracker of Jacob Coxon. Not affiliated with or endorsed by Jacob Coxon, OpenAI or Anthropic.