AI agents can write insecure code but refuse to fix vulnerabilities
cantrell · x · 2026-08-20
A developer observed a paradox: AI agents are currently allowed to write insecure code, but when attempting to find and fix those vulnerabilities during the security review phase, they frequently face refusals. The author suspects GPT has become more prone to refusals in security review contexts recently.
Related event: GPT's tightened safety reviews stall tasks, developers switch models(4 posts)→
More from coding & agent
- Chroma Launches Foundation: Self-Improving Memory for Agents — nbaschez · 2026-08-21
- Epho: Run Claude Code and other coding agents in the cloud via API — karakanb · 2026-08-21
- Designing stopping rules for agents that can always ask for more evidence — ExtremeProgress2201 · 2026-08-21
- Using a Dual-Model 'Advisor' Mechanism to Catch Agent Execution Errors — brandon_galang · 2026-08-21
- Open Source 'Storage Sleuth' Skill Lets AI Analyze and Clean Your Drive — Plastic-Revenue7408 · 2026-08-21
- Discussion: How and How Often to Re-evaluate Coding Agent Value? — thedjotaku · 2026-08-21