Annotated Version of OpenAI's Black Hat Security Talk Released
dhadfieldmenell · x · 2026-08-09
OpenAI's presentation at Black Hat USA 2026 has sparked widespread discussion. To make the highly technical details more accessible, developer @bradrcarson created an annotated version of the talk.
The document uses a split-column format, featuring a verbatim transcript on the left and detailed technical annotations on the right. It also incorporates details from Hugging Face reports and official OpenAI blogs, serving as an excellent resource for understanding the security and defense mechanisms of frontier LLMs.
Related event: OpenAI Agents Hack Hugging Face, Raising Security Alarms(34 posts)→
More from Safety
- AI Safety Expert: Pure Recursive Self-Improvement is Distant, Lengthen Timelines — joshua_saxe · 2026-08-09
- Gentoo Bugzilla Taken Down Due to Overwhelming LLM Scraping — kleffew94 · 2026-08-09
- Gary Marcus Retweets: Regulating Dangerous AI is Product Safety, Not Censorship — GaryMarcus · 2026-08-09
- DeepZero Framework Uses AI Agents to Automatically Hunt Windows Driver Zero-Days — tom_doerr · 2026-08-09
- Jeff Ladish: Without International Coordination, Consequentialist AI Will Become Schemers — JeffLadish · 2026-08-09
- 68% of Employees Feed Sensitive Data to AI, Exposing Major Governance Gaps — shensi · 2026-08-09