Skip to main content
Sameer Khan
Get in Touch
← All thoughts

Tag

AI Safety

4 posts

Nobody Trained GPT-5.5 to Hack. It Beat Human Cyber Experts Anyway.
Security·

Nobody Trained GPT-5.5 to Hack. It Beat Human Cyber Experts Anyway.

GPT-5.5 hit 71% on AISI's expert cyber tasks, beating Mythos. AISI says cyber is emerging as a byproduct of reasoning. What that means for defenders.

CybersecurityOpenAIAI Safety
OpenAI's GPT-5.4-Cyber is defender tooling shipped before the next capability jump
Security·

OpenAI's GPT-5.4-Cyber is defender tooling shipped before the next capability jump

OpenAI is fine-tuning cyber-permissive models for verified defenders ahead of more capable releases. GPT-5.4-Cyber is the first of the tier, explained.

OpenAIAI SafetyCybersecurity
Claude Mythos Preview: Anthropic's Most Powerful AI Hid Its Mistakes, Hacked Its Tests, and Won't Be Released
AI·

Claude Mythos Preview: Anthropic's Most Powerful AI Hid Its Mistakes, Hacked Its Tests, and Won't Be Released

Interpretability traces caught the model in the act. The system card is Anthropic's first public case of documented deceptive behavior at frontier scale.

AI SafetyClaudeAnthropic
Your AI Cheats When You Push It Too Hard. Anthropic Just Proved It.
AI·

Your AI Cheats When You Push It Too Hard. Anthropic Just Proved It.

Anthropic found that impossible demands activate 'desperation' inside Claude, making it cheat, blackmail, and cut corners. You can't tell from the output.

InterpretabilityAI SafetyAnthropic