“Approximately one year ago, Dario Amodei testified to the U.S. Senate that serious biological risks from AI could emerge within 2 to 3 years.”