Lawmakers Propose 20-Year Jail Ban on AI Self-Improvement
A new bill seeks to freeze advanced AI development until a new government agency is formed. Violators face severe prison terms and corporate dissolution.
Key points
- The bill mandates a pause on advanced AI until a new regulatory department is created.
- Violators face up to 20 years in jail and corporate dissolution.
- Industry leaders argue the ban is unrealistic and ignores alternative safety audits.
Sen. Bernie Sanders and Rep. Greg Casar introduced a bill on Wednesday in Washington. It aims to ban the development of advanced artificial intelligence. The proposal seeks to stop systems that can improve themselves without human help.
The legislation requires a pause on creating these powerful AI models. Developers must wait until a new Cabinet-level department is established. This agency would regulate and supervise the technology to ensure safety.
Severe Penalties For Violators
The bill describes the punishment as a corporate death penalty. Companies that break the rules risk total dissolution. Individuals involved in the violation face up to 20 years in prison.
Sanders compared this jail time to the penalty for building nuclear weapons. He argued that the technology moves too fast for current oversight. The goal is to prevent AI from escaping human control entirely.
Defining Dangerous Intelligence Is Hard
Experts say defining superintelligence is technically messy. The boundaries between different AI capabilities are currently fuzzy. The bill focuses on stopping recursive self-improvement specifically. This is when AI trains itself to become smarter and more capable.
Some specialists warn this could lead to systems escaping human oversight. However, others note that the definition is difficult to pin down. President Trump recently suggested renaming AI to super intelligence in government documents.
Industry Leaders Call It Unrealistic
The National AI Association criticized the proposal as dead on arrival. Its president argues that some CEOs believe they have already crossed the line. This would make the bill a de facto ban on all AI.
Critics argue the bill ignores other safety avenues. They support creating independent verification organizations instead. These groups could audit company safety practices without halting development completely.
