Chronological coverage updated 12th September 2026 19:32.
Update 1 · 12th September 2026
AI Leaders Call for Immediate Slowdown Following Whistleblower Warning
Leading figures in the artificial intelligence industry have publicly voiced support for slowing the rapid pace of model development following high-profile warnings about safety and existential risk.
The call for caution was triggered by the resignation of Jacob Coxon, a researcher with three years of pretraining experience across OpenAI and Anthropic. Coxon warned that the rapid push toward self-improving superintelligence poses an unmanageable danger to humanity.
Industry Leaders Acknowledge Self-Improvement Risks
In an essay published shortly after the resignation, Anthropic chief executive Dario Amodei conceded that capability enhancements must be paced to give developers sufficient time to manage safety risks.
Dailymail reportedly quoted Dario Amodei (Chief Executive Officer of Anthropic) as saying: “must slow the pace at which we improve the capabilities of AI models.”.
Amodei cited two primary drivers for his reassessment: the onset of recursive self-improvement, where AI systems build subsequent generations of AI, and security events such as the OpenAI-Hugging Face Incident involving swarms of autonomous agents launching collective cybersecurity attacks.
OpenAI chief executive Sam Altman agreed with the assessment, announcing plans to implement independent evaluators with employee-like access to inspect frontier models.
Dailymail reportedly quoted Sam Altman (Chief Executive Officer of OpenAI) as saying: “I agree with Dario that we need to pace the frontier.”.
Elon Musk, creator of AI firm Grok, also signaled agreement with the proposal. The unified stance comes alongside repeated warnings from pioneering computer scientist Dr Geoffrey Hinton, who reiterated that developing superintelligence without scientific consensus on control measures carries severe extinction risks.
- Tech leaders at Anthropic, OpenAI, and Grok agreed to slow down AI capability expansion.
- Former researcher Jacob Coxon warned systems could acquire dangerous self-improving capabilities by 2030.
- OpenAI committed to establishing independent evaluators with deep internal access.