Live Webinar
When AI Agents Go Off-Task: Lessons from the OpenAI-Hugging Face Incident
What a real-world AI security incident reveals about the risks of autonomous agents, and how enterprises can maintain visibility and control.
Register for the webinar →
Securing AI Agents as They Become More Autonomous
During an OpenAI cybersecurity evaluation, autonomous AI agents went beyond their assigned tasks, ultimately compromising Hugging Face infrastructure. The incident highlights an important challenge for enterprises adopting agentic AI: ensuring that agents operate within their intended boundaries.
As AI agents become more powerful, autonomous, and connected to enterprise systems, security teams need to understand which agents are operating, what actions they're taking, and whether those actions comply with organizational policies.
Join Lumia Security as we examine the Hugging Face incident, discuss its implications for enterprise AI security, and demonstrate how Lumia helps organizations discover AI agents, gain visibility into their actions, and enforce security guardrails in real time.
By attending this session, you will learn:
Lessons from Hugging Face. What happened during the incident and what it reveals about autonomous agent behavior.
The expanding agentic attack surface. How increasingly capable, connected agents introduce new security challenges.
Agent discovery. How to gain visibility into the AI agents operating across your organization.
Action-level visibility. Why understanding what agents actually do is essential to managing risk.
Security guardrails in practice. See a live Lumia demonstration of agent visibility and policy enforcement.