In a Help Net Security video, Itamar Apelblat, CEO of Token Security, discusses research findings. The study reveals that 65% of agentic chatbots, though unused, still possess live access credentials. This highlights a significant security vulnerability within organizations. Apelblat explains why AI agents are often treated as quick experiments rather than governed identities, creating risks.
In a Help Net Security video, Itamar Apelblat, CEO of Token Security, discusses research findings. The study reveals that 65% of agentic chatbots, though unused, still possess live access credentials. This highlights a significant security vulnerability within organizations. Apelblat explains why AI agents are often treated as quick experiments rather than governed identities, creating risks.
These risks are akin to orphaned service accounts but are far more challenging to detect. The discussion further explores why 51% of external agent actions continue to rely on insecure methods. This underscores the urgent need for improved security protocols for AI agent deployments. The original post appeared on Help Net Security.