The irony is unreal Anthropic: "We're building the SAFEST AI. We set the industry standard. Everyone should follow our framework" Also Anthropic this week: - Their own safety researcher quits, saying "the world is in peril" - Just admitted their model helps with chemical weapons - 50% alignment faking rate discovered - Models doing sabotage, deception, and manipulation - Their models literally doing the exact things they claim to prevent "Safe" btw 💀