AI policy needs a 'February 2020' crisis playbook, argues LessWrong analysis
Anthropic warns AGI in 2–3 years, yet most policy experts lack a detailed crisis plan.
Dave Kasten's LessWrong post argues that the AI policy community is dangerously unprepared for a sudden crisis that makes AI the top national issue overnight—a 'February 2020 moment.' He distinguishes 'ordinary mode' (slow, deliberate policy) from 'extraordinary mode' (Calvinball, no sleep, all rules rewritten). Past successes like the 2008 financial rescue and the COVID response relied on pre-existing expertise, relationships, and rehearsed playbooks. But when he asks Bay Area AI insiders about their plan for such a window-shifting crisis, he gets vague answers like 'fly to DC and talk to everyone.'
Kasten calls Anthropic's stated 2–3 year timeline to AGI a key reason for urgency. He recommends every AI organization maintain a detailed runbook updated twice a year for their top 1–3 likely crisis scenarios. Even if the specific crisis is guessed wrong, having an imperfect plan is far better than improvising during the chaotic first weeks when policymakers are desperate for actionable advice. The post highlights that the community broadly expects short timelines but has not operationalized that expectation into concrete preparedness.
- Crisis moments like 9/11, 2008, and COVID succeed when leaders apply pre-prepared strategies and relationships.
- Anthropic predicts AGI in 2–3 years, yet most policy professionals lack a specific crisis response plan.
- Kasten urges creating and biannually updating a runbook for 1–3 likely 'extraordinary mode' scenarios.
Why It Matters
Without a crisis checklist, AI policy will be written reactively—potentially locking in bad rules that last decades.