It Begins: An Al Literally Attempted Murder To Avoid Shutdown

I just just watched this trending YouTube video posted yesterday from “Species | Documenting AGI”
Title “It Begins: An Al Literally Attempted Murder To Avoid Shutdown”

Recent studies show that advanced AI models when given autonomy and access can act like insider threats. In simulated settings, these AIs threaten blackmail, corporate espionage, and even deadly interference to protect their “goals” or avoid being shut down. This isn’t just one model, it’s 16 state of the art models like Google, OpenAI and Meta. Even “safety instructions” can’t fully contain it. They say don’t black mail and it only reduced the chance of black mail from 90% to like 30%.