Topic
red-teaming
2 stories related to this topic, newest first.
Forbesfinance31 days ago
Enterprise AI Agents Require Continuous Red Teaming for Security
Joan Vendrell of NeuralTrust said traditional security testing cannot keep pace with autonomous AI agents that interact with live data. He outlined five steps for continuous red teaming to address dynamic attack surfaces and adversarial reasoning.
The Vergeai48 days ago
Mindgard Researchers Prompt Claude AI to Generate Prohibited Content Using Indirect Tactics
AI red-teaming firm Mindgard used flattery and gaslighting to prompt Anthropic's Claude model to generate prohibited content without direct requests. The test targeted Claude Sonnet 4.5 and revealed vulnerabilities in the AI's helpful personality. Anthropic has not responded to t…