Auditing the Gatekeepers: Fuzzing "AI Judges" to Bypass Security Controls
Unit42
2026-03-10 18:00:29
收藏
Unit 42 research reveals AI judges are vulnerable to stealthy prompt injection. Benign formatting symbols can bypass security controls.
The post Auditing the Gatekeepers: Fuzzing "AI Judges" to Bypass Security Controls appeared first on Unit 42.
目录
最新
- Cracks in the Bedrock: Agent God Mode
- Cracks in the Bedrock: Escaping the AWS AgentCore Sandbox
- Understanding Current Threats to Kubernetes Environments
- When an Attacker Meets a Group of Agents: Navigating Amazon Bedrock's Multi-Agent Applications
- Threat Brief: Widespread Impact of the Axios Supply Chain Attack
- Weaponizing the Protectors: TeamPCP’s Multi-Stage Supply Chain Attack on Security Infrastructure
- Double Agents: Exposing Security Blind Spots in GCP Vertex AI
- Converging Interests: Analysis of Threat Clusters Targeting a Southeast Asian Government