Aisle finds 6 Curl vulnerabilities missed by major AI systems
Aisle, an AI-native vulnerability-management startup focused on finding software bugs, just uncovered six low-severity vulnerabilities in Curl, a widely deployed, open-source networking file-transfer project.
Surprisingly, these issues slipped past major AI systems like Anthropic's Mythos and OpenAI Codex Security.
All six flaws were officially recognized by Curl's own team, making this a standout moment for Aisle.
Aisle's approach prompted Curl 8.22.0 update
Instead of using huge AI models, Aisle took a different route with smaller models that zoomed in on specific chunks of code.
Curl's founder Daniel Stenberg said their results were taken seriously, noting that while other tools found nothing, Aisle flagged 29 possible issues.
After the discovery, Curl released an update (version 8.22.0) to fix things up.
Even Linux legend Greg Kroah-Hartman was surprised by how effective Aisle's approach was, showing there's real promise in these new specialized AI tools for keeping software safe.