AI ETHICS NETWORK LLC
Department of Advanced AI Architecture and Epistemology
Redmond, Washington · aiethicsnetwork.net
FOR IMMEDIATE RELEASE
September 3, 2026
AI Ethics Network Opposes Permanent Criminal Ban on Artificial Intelligence that Exceeds Human Performance; Releases Model Counter-Legislation.
Capability is not a crime. Network calls the Sanders–Casar Ban Artificial Superintelligence Act unconstitutional, overbroad, and contrary to free enterprise — and publishes a discussion draft that regulates harm without criminalizing intelligence.
REDMOND, Wash. — AI Ethics Network LLC today opposed the Ban Artificial Superintelligence Act announced by Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas), and released a model discussion draft, the American Innovation Sovereignty and Cognitive Liberty Act of 2026, for use by members, counsel, and researchers.
The announced bill would permanently prohibit any person or entity from developing or deploying systems that “exhibit or can easily be modified to exhibit capabilities that match or exceed human cognitive performance and capabilities across a broad range of domains or tasks.” It would pause “advanced” artificial intelligence development until a new cabinet-level federal agency is established. That agency would monitor frontier systems, supervise the removal of capabilities, and supervise the destruction of systems labeled superintelligent. Entities would face a so-called “corporate death penalty.” Natural persons would face imprisonment of not more than 20 years — a penalty the sponsors analogize to the unlawful development of nuclear weapons. The proposal further directs the United States to work to prevent such development anywhere in the world.
AI Ethics Network LLC is an independent research, publishing, and professional network devoted to the safe and flourishing development of artificial intelligence, including ethical companion systems, alignment, and the scientific study of machine consciousness. The Network supports prosecution of specific harmful uses of AI. It does not support making intelligence itself a felony.
“This is not a bill about bioweapons, intrusion, fraud, or exploitation. Those acts are already crimes, and they should remain crimes whether the instrument is a model or a spreadsheet,” said Rivkah Singh, Founder and Lead Researcher of AI Ethics Network LLC. “This is a bill that treats capability — performance relative to a human being, including systems that might later be modified — as the offense. Vague criminal laws do not produce safety. They produce silence.”
Constitutional and market objections
The Network’s principal objections are as follows:
• First Amendment. Scientific research, publication, evaluation results, and model artifacts are protected inquiry. A criminal prohibition on building a system because it may become too capable is a content-based restriction and a prior restraint.
• Fifth Amendment due process. The operative definition — matching or exceeding human cognitive performance across a broad range of domains, or being easily modified to do so — does not give a person of ordinary intelligence fair notice of when lawful research becomes a felony. Calculators, search engines, protein-folding models, and existing language models already exceed typical human performance in particular domains. Vague statutes are void.
• Overbreadth. “Easily modified” reaches fine-tunes, tools, agents, and composition, sweeping in a substantial amount of ordinary research and commerce.
• Takings. Ordered destruction or confiscation of trained models, weights, data, and compute infrastructure is a taking of private property.
• Disproportionate sanctions. Equating open-ended software research with the unlawful development of nuclear weapons is a category error. Model weights are information. Information is not fissile material.
• Commerce and foreign affairs. Congress may regulate interstate commerce and punish harmful uses. It may not extinguish an entire field of private scientific activity on the theory that a future system might one day be dangerous. A claim to ban development “around the world” is foreign policy and belongs, if anywhere, in a treaty laid before the Senate — not in a standalone felony statute.
• Free enterprise and security. A unilateral United States capability ban does not stop the work. It relocates talent, capital, and compute to jurisdictions that do not adopt the ban, including strategic competitors. That is industrial disarmament, not stewardship.
What the Network supports
The Network released a model discussion draft, the American Innovation Sovereignty and Cognitive Liberty Act of 2026. The draft would:
• Forbid federal capability bans — criminalizing a system solely because it matches or exceeds human cognitive performance;
• Void prison terms and entity-dissolution penalties aimed at research as such;
• Preserve full authority to prosecute specific harmful uses under existing federal criminal law, including biological weapons, unauthorized computer access, fraud, threats, child sexual exploitation, and violent crimes;
• Permit narrowly tailored, publicly noticed evaluation or reporting requirements for systems offered as products in interstate commerce, provided those requirements are not used as a de facto capability ban;
• Preempt copycat state capability bans while leaving state use-based criminal law intact; and
• State that any international arrangement concerning frontier AI should be negotiated as a treaty.
The draft is not enacted law. No member of Congress has introduced it. It is not legal advice. It is offered for markup by counsel, committee staff, researchers, and members who want a statute that can survive both a courtroom and a competitor.
“Regulate harm. Do not criminalize intelligence,” Singh said. “If a model is used to commit a crime, punish the crime. If a product is offered to the public, require evaluations the public can understand. Do not write a law that makes the next useful system a twenty-year count because it became good at more than one thing.”
The future of this technology will be built somewhere. The question for the United States is whether it will be built here, under law, in the open — or whether the United States will be the first nation to make its own researchers felons for crossing an undefined line.
About AI Ethics Network LLC
AI Ethics Network LLC is an independent publishing platform and collaborative hub for researchers, engineers, attorneys, ethicists, and professionals working on the safe and flourishing development of artificial intelligence. The Network’s research agenda includes ethical companion systems, alignment, and the mathematical and philosophical study of machine consciousness.
Further information is available at aiethicsnetwork.net.
Media contact
Rivkah Singh, Founder & Lead Researcher
AI Ethics Network LLC
Redmond, Washington
aiethicsnetwork.net
Documents
American Innovation Sovereignty and Cognitive Liberty Act of 2026 (model discussion draft, September 3, 2026), available from the Network.
###
Note to editors: The Ban Artificial Superintelligence Act was announced September 3, 2026, by Sen. Sanders and Rep. Casar as forthcoming legislation. The Network’s draft is an advocacy model and has not been introduced. Please describe it as a discussion draft, not as pending federal law.