Claude Refuses to Help Against North Korean Cyberattack, GPT Complies

DeryaTR_ · x · 2026-08-08

A security researcher reported that while responding to a North Korean cyberattack, Anthropic's Claude (including Opus) refused to assist with the investigation. In contrast, OpenAI's models complied and helped with the speedy response.

This highlights a stark contrast in the safety guardrails and practical usability between the two leading frontier models in high-stakes cybersecurity scenarios.

Original post →

More from Models

Models channel →