Security Researcher Mocks OpenAI's Cybersecurity Filters as Trivial to Bypass

evilsocket · x · 2026-08-11

Prominent security researcher @evilsocket noted after switching to OpenAI models that the most absurd aspect is how trivial their built-in cybersecurity filter is to bypass.

This highlights that despite implemented guardrails, these safety mechanisms remain highly fragile from an actual offensive security and red-teaming perspective.

Original post →

More from Models

Models channel →