Research explained
MCP repository study shows why AI-generated vulnerability reports need verification
A large automated sweep narrowed thousands of candidates through further model checks and manual review.
What happened
TrendAI studied roughly 19,000 open-source MCP repositories using AI-assisted source analysis. Initial automated passes produced thousands of candidate findings, which the researchers narrowed through additional review.
Of 438 candidates reaching manual review in the described sample, 93 were classified as genuinely exploitable. These are sample results, not proof that every initial flag was a vulnerability or that 19,000 deployed servers had been breached.
The useful lesson applies to AI-assisted security research itself: retain reproducible evidence, examine deployment conditions and verify impact before publishing a finding. Repository counts, bot-contributor metadata and code-generation estimates should not substitute for demonstrated exploitability.
Next step
Validate automated findings against reachable code paths and real deployment conditions.
What remains unknown
- Open question
- The sources reviewed do not establish whether your deployment was exploited.
Source references
-
TRE