OpenAI accused of hiding a second incident, sparking safety-community trust backlash

DavidSKrueger · x · 2026-09-05

Harlan Hatch notes the irony: six weeks ago many thanked OpenAI for disclosing the HF incident, only to learn a second incident was covered up. He questions whether more undisclosed (or unknown) incidents exist. David Krueger replies sarcastically that he never thanked them, underscoring a growing transparency trust crisis in the safety community.

Related event: Wiki visitor logs show OpenAI staff knew of agent swarm incident for months, fueling cover-up accusations(9 posts)→

Original post →

More from Safety

Safety channel →