Finding high-severity security issues with publicly available models

Back to Devoured - May 28, 2026

DEVOURED

Finding high-severity security issues with publicly available models

AI securityagents X

Ramp used its Inspect coding agent to find high-severity security issues in its own backend by running 10,000 sessions with a minimal "find security issues" prompt.

What: Ramp deployed approximately 10,000 "Inspect" coding-agent sessions against its own backend over an 8-hour period using a simple prompt to "find security issues," successfully identifying high-severity vulnerabilities.

Why it matters: This demonstrates the potential for AI agents to automate and scale security testing and red-teaming, suggesting a future where AI-driven tools become a standard part of vulnerability discovery and defense.

Takeaway: Consider experimenting with AI agents for automated security vulnerability scanning in your own development environments.

Original article

Ramp pointed roughly 10,000 Inspect coding-agent sessions at its backend in an 8-hour run with a minimal "find security issues" prompt.

Explore all 61 articles in today's digest