Research · Thursday 15 October 2026 · story 2
Benchmark finds coding agents often fail secure code tasks
Hugging Face reports that the Last Secure Code Benchmark tests secure code generation with 450 tasks from 150 real vulnerabilities across seven languages and three vibe-coding scenarios. It says six frontier agents passed 73.6% to 91.3% functionally, but Claude Fable 5 reached 25.1% for functional and secure code.
Why it matters. Teams using coding agents should add security checks, because even Claude Fable 5 was functional and secure on only 25.1% of tasks.
Read the original at huggingface.co