Dev grapples with Codex trust and security: all code is basically insecure

code_star · x · 2026-09-18

codestar offers a real-time reflection (not a critique of OpenAI's or Anthropic's security competence) on what agentic coding means for security: he uses and loves Codex, but must now think hard about relying on it if he can't trust its availability or data security against malicious parties. His take: all code is basically insecure, and even 24/7 model-driven patching won't fix that—security is risk mitigation and deterrence, and the bigger the prize, the harder the defense.

Original post →

More from AGI Musings

AGI Musings channel →