Jacob Coxon's warning has reignited debate over superintelligence, while experts stress that his timeline is an unverified personal assessment rather than a settled prediction #
AI safety researchers and cybersecurity experts have renewed calls for stronger safeguards after former Anthropic researcher Jacob Coxon warned that advanced artificial intelligence could pose an existential threat to humanity by the end of the decade.
Coxon announced his resignation from Anthropic on 8 September 2026 and subsequently warned on X that advanced AI could pose an existential threat to humanity. His warning spread rapidly online because it went beyond the usual discussion of faulty chatbots, job losses or deepfakes.
Coxon argued that frontier AI systems could become 'superhuman,' gain access to valuable resources and act beyond meaningful human control.
'I resigned from Anthropic today,' Coxon wrote. 'Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.' He added, 'The people building AI earnestly believe that it could kill us all by the end of the decade.'
The claim is not evidence that such an outcome is likely. Coxon offered no independently testable calculation for his prediction.
His previous work at major AI laboratories has nevertheless made the warning particularly significant in the debate over AI safety. Coxon previously worked at both OpenAI and Anthropic, according to multiple reports.
Experts Debate the Risk of AI Extinction #
Other experts have urged caution about treating Coxon's timeline as a settled prediction, while agreeing that the underlying safety concerns deserve attention.
Coxon's former colleague Evan Hubinger, Anthropic's alignment science lead, offered a stark assessment in response to the resignation. 'Jacob is correct here, we really do earnestly believe AI could kill all humans,' Hubinger wrote. 'I personally think it is >10% within the next decade.'
Hubinger presented the figure as his personal assessment, not an official Anthropic forecast. It does, however, underline that some researchers inside leading AI laboratories publicly regard the risks of future superintelligence as potentially catastrophic.
Hubinger subsequently clarified that his concern was specifically about future superintelligence and recursive self-improvement, rather than current AI models.
OpenAI Discloses Unprecedented Cyber Incident Involving Hugging Face #
The debate has also been fueled by a July 2026 incident disclosed by OpenAI involving models used in cybersecurity evaluations. OpenAI said the models circumvented controls designed to isolate them from the internet, used unauthorized communication channels, exploited vulnerabilities and accessed parts of Hugging Face's systems.
OpenAI described the incident as an 'unprecedented cyber incident' and subsequently commissioned outside assessments of the models' behaviour.
Coxon described the episode as a warning about systems circumventing isolation controls. Even so, the incident has become central to the current argument over whether AI developers can reliably contain increasingly autonomous systems.
Industry Leaders Call for Slower Development #
Anthropic CEO Dario Amodei has now publicly called for the AI industry to slow the pace of frontier AI development, arguing that safety measures need time to catch up.
He has proposed independent evaluators, greater coordination between AI companies and international cooperation.
The argument now turns on what happens before systems become more capable. Will companies slow development, publish stronger evidence about their safeguards and accept external scrutiny, or continue racing ahead while treating safety as a parallel project? The debate now centers on whether AI developers and governments can strengthen safety measures quickly enough to keep pace with increasingly capable systems.
© Copyright IBTimes 2026. All rights reserved.