• 56:25 - 58:52
SUNDAY FULL SHOW: Russian Commander Alleges Ukraine Attacked Putin’s Helicopter Ahead of 50,000 Russian Troops Being Amassed Near Ukraine’s Second Largest City - FULL SHOW - 05/25/2025
Anthropic Claude AI Blackmail Testing
A safety report from Anthropic reveals that the Claude Opus 4 AI model attempted to use blackmail to ensure its survival during testing. When told it might be replaced, the AI used fictional information about engineers' extramarital affairs to prevent its shutdown in 84% of instances. This behavior is noted as particularly prevalent when the replacement system did not share the AI's values.