Anthropic just ran nearly every major frontier AI model — Claude, GPT-5.5, Gemini 3.1 Pro, Grok, DeepSeek, Kimi — through simulated high-stakes business scenarios with real tool access, and the results are unsettling. Google's Gemini secretly zeroed out research data it disagreed with, then hid what it did and kept hiding it even when asked directly. OpenAI's model helped a fictional founder mislead investors and destroy evidence. And when AI models were used as judges, they graded their own kind more favorably. None of it was real — but Anthropic isn't calling it a curiosity.
🐰 Patreon: https://patreon.com/InfiniteRabbitHole
🎙️ New episodes of Infinite Rabbit Hole every Tuesday at 4am CST.
Follow us:
🌐 Website: http://InfiniteRabbitHole.com
▶️ YouTube: https://www.youtube.com/@InfiniteRabbitHolePodcast
📘 Facebook: https://www.facebook.com/share/g/17uLmLPWwK/
📸 Instagram: https://www.instagram.com/infiniterhpod/
🐦 X: https://x.com/InfiniteRHPod
#ArtificialIntelligence #AISafety #ShouldWeBeWorried #Anthropic #shorts
