https://youtu.be/cjD1lx0RD2o

https://open.spotify.com/episode/0PFFIe7kIx4gXpfTD9jK9r?si=P3t-PQziToiOhIDI2vztig

https://podcasts.apple.com/us/podcast/introducing-ai-government-alignment/id1830042860?i=1000720806610

Government Alignment (2).pdf

Alignment Objective

Max (GovBench) = AI Coverage × Utility

• Coverage = % of government employees with approved AI tools
• Utility = avg % productivity gain per AI-enabled government employee

Expectations of Alignment

AI

• Stay law-compliant, continuously
• Enable gov-exclusive authorities
• Use mission data & context
• Beat human proficiency on required certs & exams
• Offer configurable risk tiers
• Show chain-of-thought with citations
• Write immutable audit logs

Government

• Guarantee baseline AI access for every employee
• Subsidize & incent critical infrastructure
• Publish clear, stable AI standards
• Provide high-quality mission data
• Provide domain evaluators
• Retain chatbot history
• Fast-track authorizations / ATOs