Published: Jun 11, 2026, 23:57 IST | Updated: Jun 11, 2026, 23:57 IST
A new assessment of 100 AI agents reveals serious security weaknesses, showing that most can be manipulated through prompt injection attacks. The study highlights how hidden instructions inside emails or documents can mislead AI systems that have access to sensitive tools and data. Experts warn that as AI agents take on more real-world tasks, security gaps could become a major risk for organisations.