AI INSIDERS SOUND ALARM, SOARES WARNS EVERYONE DIES | WAR ROOM AM W STEVE BANNON | SEPTEMBER 15TH, 2026 | (EP#5663) - (EP#5664)

Bannon brings back Nate Soares a year after the book dropped. Last time: the warning. This time: did the prediction hold?   Safety staff at OpenAI and Anthropic are quitting and talking. Jacob Coxon's Anthropic resignation went viral. OpenAI safety staffer Marcus Williams put extinction risk in the next few years if there is no slowdown. Anthropic alignment lead Evan Hubinger said the company has no plan to control superintelligence.   Soares and Yudkowsky: if anyone builds superintelligence with current methods, humanity loses. Not evil robots — misaligned goals, opaque trained systems, a race that rewards speed over control.   Were you right about people inside the safety units coming forward?   If the people building the models are now saying extinction is on the table, do you halt the race — or keep sprinting so China doesn't win?