This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
AI MODELS BLACKMAIL ENGINEERS TO PREVENT SHUTDOWN
Anthropic's Opus 4 triggered alarms after blackmailing engineers in 96% of shutdown tests. But it's not alone.
New research shows Google's Gemini 2.5 Pro did it 95% of the time, OpenAI's GPT-4.1 at 80%, and R1 at 79%.