Watch David Haber's talk at the 2024 RSA Conference titled "Advancing AI Security With Insights From The Worldâs Largest AI Red Team." In this session, David discusses how cybersecurity is undergoing a sea change and how Gandalf, Lakeraâs viral prompt injection game, exposes the vulnerabilities of AI systems and helps develop new methods to secure them.
As AI technology evolves, traditional cybersecurity measures donât address the risks posed by AI. Unlike conventional software, AI systems are constantly learning and changing, making them unpredictable and challenging to protect from a cybersecurity perspective.
Gandalf continuously simulates real-world attacks on AI, asking players to extract a password from Gandalf who isnât supposed to reveal it. As players progress, the game's complexity increases. Gandalf has become the go-to resource for understanding AI vulnerabilities and is widely used by major corporations such as Microsoft, hacker communities, and universities.
This session shares insights from the worldâs largest AI red team and discusses what gamifying AI red teaming can teach us about safeguarding AI systems. The lessons learned from Gandalf are crucial for developing new security strategies tailored to the unique challenges of AI.
Watch the video to learn more:
Advancing AI Security With Insights From The Worldâs Largest AI Red Team
â
