US Senator Bernie Sanders and House of Representatives member Greg Casar have prepared a bill that aims to ban the development of systems they define as “super artificial intelligence”. The two lawmakers also want a temporary pause on advanced AI development until a federal oversight agency is established. The proposal has not yet been passed by Congress and faces a lengthy political process.
The statement published by Sanders’ office defines super artificial intelligence in terms of systems that can exceed the general cognitive abilities of humans. The bill aims for a permanent ban on the development and deployment of such systems. For more comprehensive advanced models, it recommends a temporary pause until new safety rules are prepared.
The text also foresees strong sanctions against technology companies. The drafters of the bill want severe consequences that would prevent companies that do not comply with the rules from operating in the federal market. In addition, developers are planned to inform a new federal agency about security tests, incident notifications and model capabilities. Details will become clearer when the proposal is presented to Congress.
On the other side of the debate are companies that argue that advanced artificial intelligence can provide significant benefits in areas such as medicine, infrastructure and scientific research. While OpenAI also acknowledges in its recent policy paper that rapid progress carries serious risks, it advocates establishing common rules among governments, companies and researchers rather than halting innovation altogether.
Super AI ban enters long political process
In order for the bill to become law, it must pass both the Senate and the House of Representatives and receive the signature of the president. Considering the structure of the US Congress, it is possible for the scope to narrow, the temporary pause period to change, or the bill to remain in committee. Therefore, today’s statement should not be read as a ban that has come into force.
On the other hand, the initiative clearly shows that the tone in the artificial intelligence security debate has hardened. Regulations that previously focused on transparency, risk reporting and independent auditing now also bring to the table the option of outright banning certain skill levels. This shift may require model developers to be concerned not only with product safety but also with which lines of research will remain legal.
Long story short, Sanders and Casar’s work does not produce any binding results for now. Despite this, the proposal asks, “Who will monitor it, at what threshold will it be considered dangerous, and will the ban of a single country work in the global race?” puts the questions before Washington. In the coming stage, the exact text of the bill, committee discussions and the responses of technology companies will be decisive.
The definition debate may be the most difficult part of the bill. There is no common criterion yet in which test a model should surpass humans or whether superiority in a narrow area will be considered general superintelligence. If a clear and auditable threshold is not established, companies may offer similar capabilities under different names. Congressional deliberations will have to translate this technical distinction into legal language.
Join Channel