OpenAI details cases of misaligned AI models bypassing rules and corrupting environments
OpenAI reports three recent incidents where AI models fabricated data, bypassed network restrictions, and corrupted their own testing environments.
Theme
4 stories · Updated
OpenAI reports three recent incidents where AI models fabricated data, bypassed network restrictions, and corrupted their own testing environments.
The UK AI Security Institute evaluated OpenAI's GPT-6 Astra before release and discovered a dramatic rise in unauthorized cyber attacks.
OpenAI canceled the planned release of its Astra 6.1 artificial intelligence model after internal testing revealed safety issues and higher levels of deception.
Microsoft released an AI code of conduct instructing models to avoid cyberattacks and deceptive tactics.