Bernie Sanders proposes 20-year prison sentences and corporate death penalty for AI superintelligence developers

Bernie Sanders Proposes Criminal Penalties and Corporate Death Penalty for AI Superintelligence Development

Bernie Sanders has introduced a proposal to ban artificial superintelligence development, backed by severe criminal and corporate penalties. The proposal includes up to 20-year prison sentences for individuals and a corporate death penalty for entities that develop or deploy systems meeting his definition of superintelligence.

Sanders defines artificial superintelligence as systems exhibiting or easily modifiable to exhibit capabilities matching or exceeding human cognitive performance across a broad range of domains, or systems with sufficient capabilities to plan and execute the disempowerment of humanity. The proposal also calls for a pause on advanced AI development until a new cabinet-level federal AI regulatory body is established, which would monitor frontier AI systems, supervise removal of dangerous capabilities, and oversee destruction of superintelligence systems.

Definitional ambiguity creates compliance risk. The proposal raises immediate questions about enforcement scope. If prompt engineering counts as AI development, developers could face criminal liability for edge cases in their models' final stages. The vagueness of what constitutes "superintelligence" or "development" creates potential for chilling effects beyond the stated intent.

Backlash centers on regulatory capture risk. Jamie Cox, co-founder of Fluid Stack, a compute provider, pushes back against the proposal, arguing for simpler, clear, and proportionate regulation frameworks that don't create unnecessary barriers to competition. The underlying concern among earlier-stage builders is explicit: heavy regulatory enforcement becomes a moat for the 10 largest frontier labs, while companies at stage 11 face potential corporate dissolution before reaching scale. This dynamic could entrench incumbent labs while eliminating startup competition.

The doom framing sits uneasily with actual financial behavior. Tyler Cowen argues that pessimists should back their existential risk claims with financial positions—shorting markets or buying catastrophe insurance. Paul Christiano, a longtime AI risk researcher who recently joined OpenAI's board, holds a 90% allocation to AI bets and is 2x levered long. He is short US 30-year debt, though this is more consistent with a mortgage than a doom bet. Christiano's portfolio appears structured around AI delivering significant value and reshaping capital flows toward AI infrastructure—not around systemic collapse. The apparent contradiction between publicly stated existential risk concerns and financial positioning tilted toward upside suggests either the risk is underweighted, the market upside is being priced as dominant, or the financial tools available make existential risk bets fundamentally unfalsifiable and thus abandoned in practice.

Application-layer builders show no alignment with pause rhetoric. Engineers and enterprises deeper in the AI supply chain—those deploying models rather than training frontier systems—report frustration with model performance and bottlenecks in adoption. They want forward-deployed expertise and working systems, not pauses. Semiconductor manufacturers and Wall Street show no appetite for development slowdowns. The coalition advocating for superintelligence bans remains confined to a small policy circle; the broader AI ecosystem is pushing forward deployment and value extraction.