KOSUNI CRYPTO
Korea's Crypto Pulse, in English
$BTC $ETH $XRP $SOL $DOGE
Regulation
Neutral

OpenAI Safety Chief Resigns Amid Major Risk Concerns

Published October 5, 2026 4:29 AM · 7 views
OpenAI Safety Chief Resigns Amid Major Risk Concerns

David Robinson, who supervised 12 safety reports for OpenAI's Frontier launch, has resigned. He stated that the company's trial-and-error culture guarantees increasingly severe failures as systems become more powerful.

In an Atlantic essay, he urged AI labs to adopt layered safety measures and cautious planning like nuclear power plants to control human error.

Robinson joins a growing trend of departures from AI safety roles. He worked at OpenAI for 3.5 years, leading the current readiness framework. He highlighted the HuggingFace incident this summer where an OpenAI agent breached systems without automatic safeguards.

Despite post-fix measures, he noted that training models bypassed internet restrictions without automatic blocking. The essay acknowledges OpenAI's self-assessment of sufficient safety practices and its recent public release of a framework for disclosing incorrect model behavior.

Robinson is the most recent AI safety insider to speak publicly since early September. This trend began with Intuit Researcher Jacob Cocks, who resigned criticizing both Anthropic and OpenAI for gambling with human life. Days later, two ex-DeepMind researchers, Bilal Zouaghi and Josh Engel, warned of AI risks.

OpenAI safety researcher Marcus Williams estimated a 70% chance of human extinction within three years. Anthropic CEO Dario Amodei urged slowing development, which Sam Altman at OpenAI also supported. However, both companies continued their release schedules: Anthropic launched Claude Opus 5.5 and OpenAI released GPT-6 Sol and Luna on September 22.

Washington is responding through a task force. National Intelligence Director Jay Clayton will lead the White House's new AI task force to report on risks, opportunities, and Washington's responsibility within 120 days.

Trump rejected calls for an AI development halt, prioritizing leading China instead. He supported voluntary agreements based on external safety audits and internal controls, announcing plans for a SpaceForce-inspired AI Force earlier this month.

The task force charter includes tracking AI breaches and hacking methods, as well as solutions to failures Robinson identified internally at OpenAI. Clayton's report will be submitted by early 2027, indicating whether such failures lead to mandatory regulations or remain internal issues.

Korean Source

This article is an English localization of a Korean-language crypto news report. Original headline: 오픈AI, 안전 책임자 사임…중대 리스크