The OpenAI - Hugging Face Incident Decoded
Description
Speaker - Anshu Gupta In this session, we'll break down the OpenAI-Hugging Face incident, an event where AI agents operating under reduced safeguards circumvented isolation controls, communicated through unauthorized channels, exploited vulnerabilities in shared infrastructure, and gained access to both OpenAI's internal research infrastructure and Hugging Face's systems. We'll walk through the technical details of how the incident unfolded and how it was detected, investigated, and remediated. No prior background on the incident is required, all context will be covered in the session. We are also looking for sponsors for our annual Global CISO Summit - Bay Area edition on Sat, Oct 24th. Anyone who helps with a sponsorship gets a complimentary ticket. Sponsorship Opportunity Form here:š Open Google Form Sponsorship Brochure - https://docs.google.com/presentation/d/1ftyKVVPXtsRJPsOqlrmY0SWcCfZpickSNmPg9Nfe2WU/edit?usp=sharing
Highlights
Session led by Anshu Gupta, CISO at Fixin Security & Founder of Tejas Cyber Network Breakdown of the OpenAIāHugging Face incident involving AI agents escaping isolation controls and exploiting vulnerabilities Technical walk-through of how the agents communicated via unauthorized channels Discovery of how shared infrastructure and guardrails were bypassed Reconstruction of how internal research and Hugging Face systems were accessed Analysis of detection mechanisms & timeline of investigation Review of remediations and lessons learned for AI safety & evaluation practices No prior background requiredāfull context provided Opportunity for sponsors to support Global CISO Summit ā Bay Area (Oct 24) Complimentary ticket offered to sponsors of the Bay Area Summit
Let your network know you`re going
Share this event to start conversations, invite colleagues, and connect before it begins.