Physix Frontier · News Briefing Card (enbrief · Nov 20, 2026)

Sanders and Casar move to ban superintelligence, current AI may fall under law

KEY FACTS

  • Casar said the new Congress will prioritize a short bill banning recursive self-improvement.
  • Sanders and Casar introduced a bill in September banning superintelligence and six categories of systems with precursor characteristics.
  • The bill sets a maximum of 20 years in prison and a 10-year industry ban for reckless violators.
  • A newly created AI Department secretary would decide on their own whether a system has precursor characteristics, with no quantitative standard.
  • A September poll showed 78% of Americans believe advanced AI could destroy humanity.

KEY DATA

78%Share of Americans who believe advanced AI could destroy humanity
20 yearsMaximum prison term for reckless violators
10 yearsIndustry ban period for violating employees
30 daysDeactivation period for systems found in violation

PHYSIX OBSERVATION

The bill treats "possibly possessing" precursor characteristics as grounds for conviction, yet it does not define recursive self-improvement and sets no quantitative threshold, which amounts to handing the judgment to a newly created department. For developers, training a model with a coding agent could cross the line, and compliance risk spreads from the lab into everyday R&D. If legislators truly want to manage risk, they should first add measurable standards rather than substitute criminal liability for technical assessment.

Source: enbrief original report ↗