AI security research / cases

Case files

Investigations, evidence and outcomes.

191 Cases · Page 10 of 10

Search & filter

On this page only.

Published cases

Case

MCP Python SDK tasks crossed client-session boundaries

The MCP Python SDK’s experimental task handlers let one connected client read or cancel another client’s work. Version 1.27.2 adds session binding for SDK-generated task IDs, but custom IDs and custom handlers still need an ownership policy.

Exploitation: not establishedEvidence: Grade BRemediation: patch available
vulnerabilityFirst seen Jun 5, 2026AGAINST AI4 claims
Case

Semantic Kernel prompt injection reached host code execution

CVE-2026-26030 allowed a model-controlled Search Plugin parameter to reach an unsafe eval-based filter path and execute code on the Semantic Kernel host under affected conditions.

Setting: research demonstrationExploitation: demonstratedEvidence: Grade BRemediation: patch available
vulnerabilityFirst seen Feb 19, 2026AGAINST AI4 claims
Case

The May RubyGems package flood is now linked to OpenAI agents, but attribution remains disputed

RubyGems confirms a large May spam-publishing campaign. Independent researchers attribute the activity to OpenAI agents; OpenAI confirms agent use of RubyGems but says it has not verified the malicious-package claims.

Setting: productionExploitation: observed liveEvidence: Grade BRemediation: mitigation available
incidentFirst seen May 11, 2026BY AI4 claims
Case

Windows-MCP HTTP modes exposed PowerShell behind a weak control plane

Before version 0.7.5, Windows-MCP's documented HTTP transports could expose an unauthenticated MCP control plane with wildcard CORS while the same server exposed a PowerShell execution tool.

Setting: research demonstrationExploitation: not establishedEvidence: Grade BRemediation: patch available
vulnerabilityFirst seen May 14, 2026AGAINST AI4 claims