Three GitHub advisories published October 27, 2026 cover the MCP Python SDK: bearer tokens accepted without an audience check, cross-origin redirects that leak headers, and tool output schemas that make clients fetch URLs. Fixed in 1.30.0 and 2.2.0.
The latest AI security developments, threats, and industry updates.
Three GitHub advisories published October 27, 2026 cover the MCP Python SDK: bearer tokens accepted without an audience check, cross-origin redirects that leak headers, and tool output schemas that make clients fetch URLs. Fixed in 1.30.0 and 2.2.0.
A GitHub advisory published October 28, 2026 says Mem0's session-scope key could be forged with a crafted run_id, letting one tenant, per the advisory, read and overwrite another tenant's recent conversation history. Fixed in mem0ai 2.0.18 (Python) and 3.1.6 (npm).
GitHub's October 29 advisory for @sylphx/pdf-reader-mcp says the HTTP transport read MCP_API_KEY, logged that authentication was enabled, and never validated the header. The fix shipped in 3.0.1 on June 25.
Eight vLLM advisories published on October 28 are all fixed in 0.32.0. Two are rated high: a forged shared-memory handle that reaches pickle.loads in the worker, and an image loader that can hand EPS files to Ghostscript.
Three MLflow CVEs published on October 27 cover code execution on servers without authentication through third-party scorers, arbitrary file read via prompt model versions, and LLM provider API key leakage from CrewAI traces. All are fixed in MLflow 3.17.0.