“An AI Self-Regulatory Organization (SRO) should focus exclusively on catastrophic risks such as cyber and CBRN (chemical, biological, radiological, nuclear), not on issues like disinformation or microaggressions.”