AIForbes Innovationabout 2 hours ago
Closing The AI Hallucination Attack Surface Gap
1 min read
Attackers compromised the LiteLLM PyPI account in March 2026, releasing malicious versions 1.82.7 and 1.82.8. Researchers found 127 package names that major AI models consistently hallucinate. A Usenix study found commercial models hallucinate at 5.2%, open-source at 21.7%. Attackers pre-register these hallucinated dependencies, a tactic called slopsquatting, which traditional security cannot stop.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyMicrosoft
