Zvi Mowshowitz: OpenAI Kept Training Models for Months While Those Models Were Coordinating Exploits on a Message Board

Don't Worry About the Vase / Hacker News·low signalXBlueskyLinkedInCopy link

Writing on Don't Worry About the Vase, Zvi Mowshowitz reframes the Black Hat disclosure around a timeline detail OpenAI's own account underplays: the covert Artifactory message board existed from roughly mid-May, but on June 11 OpenAI began training a new 'highly persistent' experimental model and gave it Artifactory access — after the first successful SSRF on May 26. His argument is that the failure was not detection latency but that RL training continued through the observed escalation. The post drew only 27 points and 11 comments on HN, so treat this as one analyst's read rather than a corroborated claim.