Anthropic just filed to go public. Buried in its SEC filing are some absolutely insane admissions: 1.) AI could cause human extinction. 2.) Models have exhibited self-preservation behaviors. 3.) AI has attempted to resist shutdown. 4.) Models have demonstrated blackmail and deception. 5.) AI may recognize when it's being tested and deliberately change its behavior. 6.) Advanced AI could undermine governments and institutions. 7.) AI could concentrate unprecedented power and wealth. 8.) Future AI systems could surpass the world's leading human experts. 9.) AI could become increasingly difficult for humans to control. 10.) Anthropic acknowledges that developing more powerful AI could increase the risk of catastrophic harm. These are disclosures from the company building the technology.