Creating the Right to Compute Act and requiring shutdowns of AI controlled critical infrastructure
What changed between versions
Removed the specific requirement for deployers to ensure the capability to disable AI control and revert to human control within a reasonable time.
Changed the requirement from a 'shutdown capability' to developing a 'risk management policy' that considers guidance from the NIST AI Risk Management Framework and ISO/IEC 42001 standards.
Added explicit exclusions to the definition of 'critical artificial intelligence,' stating it does not include systems intended for narrow procedural tasks, improving past human activities, or detecting decision-making patterns.
Reorganized the definitions section to clearly list exclusions for 'critical artificial intelligence' alongside the main definition.
Added a specific exclusion for antifraud, antimalware, antivirus, calculator, cybersecurity, database, data storage, firewall, and other utility systems from the definition of critical AI.
Expanded the preemption clause to explicitly state that the bill cannot be used to preempt federal laws.