Chloé Messdaghi & Kazimir Schulz: Challenges of Responsible Disclosure in AI Systems
A DEF CON 32 talk where researchers Chloé Messdaghi and Kazimir Schulz discuss the difficulties of responsible vulnerability disclosure in AI systems.
The talk examines issues such as attacks on AI models, data manipulation, and exploitation of model bias.
Using real-world examples, the speakers explain how disclosure processes can be improved to strengthen AI security and increase trust in these systems.
The talk examines issues such as attacks on AI models, data manipulation, and exploitation of model bias.
Using real-world examples, the speakers explain how disclosure processes can be improved to strengthen AI security and increase trust in these systems.