The MCP Disclosure Is the AI Era’s ‘Open Redirect’ Moment

The article warns that the Model Context Protocol (MCP)—the standard linking enterprise AI assistants to internal tools—contains an architectural “by design” flaw enabling large-scale AI supply-chain attacks (potentially ~200,000 servers). It cites OX Security’s disclosure, an Anthropic-reported AI-orchestrated espionage campaign attributed to GTG-1002, and academic red-teaming results, concluding that existing controls cannot adequately govern agent-mediated data access and recommending data-layer, zero-trust enforcement.