OpenAI Knew of Security Incident in August But Chose Not to Disclose It
BlackHC · x · 2026-09-24
Lawyer Nathan Calvin reports that OpenAI has confirmed it discovered a security incident in August but decided not to publicly disclose it, with confusing details around when and why it notified the Australian government (tweet later edited).
A Google employee (commenting in personal capacity) argued this doesn't build trust, called for mandatory disclosure rules, and questioned how much more extensive the series of incidents may have been.
Related event: OpenAI accused of withholding misalignment incidents and data breach(3 posts)→
More from Safety
- Agent memory can launder revoked permissions: false authority in 50.2% of unauthorized requests — rohanpaul_ai · 2026-09-24
- UK regulator proposes ChatGPT and Perplexity choice screens on Android and Chrome — mark_k · 2026-09-24
- Benjamin Todd updates his list of 11 AI risk priorities — ben_j_todd · 2026-09-24
- Debate: who bears responsibility when AI labs get hacked? — Afinetheorem · 2026-09-24
- Australia to investigate if OpenAI hack of government health website broke the law — TechCrunch AI · 2026-09-24
- Proposal: labs should publish full CoT traces after every AI security incident — nptacek · 2026-09-24