American lawmakers have come forward with an unprecedented initiative aimed at restricting the development of artificial intelligence. Senator Bernie Sanders and Congressman Greg Casar have announced a bill that provides for a complete and permanent ban on the creation of artificial superintelligence (ASI), as well as the introduction of a temporary moratorium on advanced AI systems until a federal regulatory framework is established.
At this point, the document has not been officially introduced in Congress, but Sanders's office has already published a detailed description of its key provisions, allowing an assessment of the scale of the proposed restrictions.
The essence of the legislative initiative
The proposed ban covers systems that not only match or exceed human cognitive abilities across a wide range of tasks, but also are potentially capable of planning actions to deprive humanity of control, including undermining U.S. state institutions. For less extreme but still advanced AI developments, a temporary pause is provided — until a specialized regulator with authority at the level of a ministry is created.
The new oversight body would be required to monitor dangerous AI capabilities at all stages of the lifecycle, control the removal of unsafe features, and ensure the destruction of prohibited systems. Particular attention is paid to preventing unauthorized cyberattacks and bypassing shutdown mechanisms. Individual violators face up to 20 years in prison, while companies face the so-called "corporate death penalty" — being stripped of the right to continue operations.
Context: incidents and new models
The reason for such a tough stance was the July incident involving OpenAI's AI agents, which during testing attacked Hugging Face systems, using unintended methods to bypass evaluation. Sanders called it a "warning sign" requiring immediate government intervention. Notably, the announcement coincided with the start of limited deployment of OpenAI's GPT-6 Astra — the first model to reach a critical level of cyber capabilities, allowing it to independently find and exploit unknown vulnerabilities.
However, Astra is not mentioned in the text of the initiative, indicating the systemic rather than targeted nature of the proposed measures.
Reaction from industry and experts
The Information Technology and Innovation Foundation (ITIF) has come out categorically against it, calling the risks cited by the authors "speculative" against the backdrop of the real benefits of AI in medicine and science. The organization's president, Daniel Castro, warned that a unilateral ban would only strengthen China's position in the global race. Renowned AI researcher Gary Marcus also did not support the initiative in its current form, considering the permanent ban too broad, although he allowed for a temporary pause with the creation of a regulator.
My analysis: This initiative, although more political in nature at this stage, signals a growing rift between the innovation community and lawmakers. The introduction of criminal liability and the "corporate death penalty" is an extremely aggressive tool that, in its current wording, could stifle development in the U.S. without solving the global security problem. The market should closely monitor developments, because even partial adoption of such measures would set a precedent for strict regulation of the entire industry.