AI Documentation Vulnerability: How Unclaimed References Threaten Corporate Networks
New research highlights a security flaw where AI agents can be tricked into installing malware by following outdated or hallucinated documentation references on corporate websites.