Tagged "model-vulnerabilities"
3 articles tagged model-vulnerabilities, 1 May 2026 to 25 May 2026. Newest first.
-
AI Guardrails Stripped From Meta and Google Models in Minutes
Security researchers demonstrate vulnerabilities allowing rapid removal of safety guidelines from commercial LLMs. Critical implications for organizations relying on guardrails in locally-deployed or fine-tuned models.
-
Claude Opus 4.7 System Prompt Leaks Raise Local Deployment Questions
Security researchers report Claude Opus 4.7 randomly leaking its system prompt, highlighting vulnerabilities in proprietary models and reinforcing the case for transparent, locally-controlled LLM deployments.
-
96.8% of MCP Tool Descriptions Don't Warn the Agent About Destructive Behaviour
A critical safety analysis of Model Context Protocol tool descriptions reveals widespread gaps in agent safety guardrails, with implications for local LLM applications using autonomous agents.