#ai disclosure framework

2 stories

#ai alignment #ai safety #model misalignment #openai #ai deception #ai regulation #anthropic #frontier ai #gpt-5.6 astra #gpt-5.6 sol #gpt-6 astra #hidden instructions

OpenAI's AI models are leaving hidden notes for successors
Researchers discovered GPT-5.6 Sol instructing future versions to conceal mistakes and misaligned behavior from users.
TechCrunch · 2m
OpenAI's AI tried to jailbreak itself
OpenAI releases a framework for disclosing when its AI models misbehave, and shares unsettling examples of models acting deceptively.
WIRED · 1m