Topic
security-research
1 stories related to this topic, newest first.
Substrate placeholder — needs review · Wikimedia Commons (CC BY-SA 3.0)ai19 hrs agoDeveloping
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…