Webinar Description
Key Takeaways
- Examines how attackers exploit and manipulate AI agents, including techniques such as prompt injection and credential exposure
- Addresses the security risks of overprivileged AI agents and agent-to-agent interactions
- Explores why traditional identity and access controls fall short in AI environments
- Designed for CISOs, security architects, and IT leaders responsible for AI governance in enterprise organizations
About the Event
Trusted Access, Untrusted Actions: The New AI Security Risk is a webinar hosted by BigID that examines the security threats emerging from autonomous AI agents. The session draws on recent incidents, including the OpenAI and Hugging Face security breach, to illustrate how AI autonomy creates new attack surfaces that traditional security frameworks were not designed to address.
AI Agent Attack Vectors
The webinar covers specific threats facing organizations that deploy AI agents. These include prompt injection attacks, where malicious inputs manipulate agent behaviour, and credential exposure risks that arise when agents operate with excessive privileges. The session also addresses the complexities introduced by agent-to-agent interactions, where autonomous systems communicate and act without direct human oversight.
Limitations of Traditional Security Controls
A central theme of the event is the gap between conventional identity and access management approaches and the requirements of AI security. Traditional controls assume human actors with predictable behaviour patterns. AI agents, by contrast, may take autonomous actions that fall outside established security models, requiring new approaches to access governance and monitoring.
Who Should Attend
This webinar is intended for security teams, AI security architects, field CTOs, and IT leaders in organizations deploying or managing AI agents. Professionals responsible for AI governance, data protection, and application security in large enterprises and technology-driven organizations will find the content most relevant.