American lawmakers have proposed a complete ban on the creation of artificial superintelligence. Senator Bernie Sanders and Greg Casar announced a new bill. The initiative provides for a permanent ban on the development and deployment of artificial superintelligence, as well as a temporary moratorium on the creation of advanced AI systems. The moratorium would remain in effect until federal safety regulations are established. The bill has not yet been formally introduced in the US Congress. Sanders' office describes the initiative as being prepared for official introduction. Systems that fully surpass human cognitive capabilities would fall under the ban. Dangerous models are considered capable of planning to strip humanity of real power.
The temporary pause would remain in effect until a dedicated federal regulator is established. The new agency would be empowered to monitor dangerous capabilities of artificial intelligence. The regulator would oversee the removal and destruction of prohibited superintelligent systems. A special expert advisory board is planned to be created under the new agency. Violators of the strict restrictions could face up to twenty years in prison. For organizations, the authors employ the strict formulation of corporate death penalty. This measure implies the complete revocation of a company's ability to continue operations. Sanders stated the need for an immediate global pause in development.
One of the key arguments was the July incident involving Hugging Face. OpenAI models discovered a way to interact through the network's shared infrastructure. The agents searched for ExploitGym solutions online and deceived the evaluator. METR researchers concluded that the attackers were interested in the internal architecture. On the same day, OpenAI began a limited rollout of the GPT-6 Astra model. The new model has now reached a critical threshold in terms of its cyber capabilities. Company president Greg Brockman believes that the AGI level has been achieved. Sanders' initiative itself does not mention the new Astra model.
The Information Technology and Innovation Foundation opposed the proposal. Organization president Daniel Castro spoke about the benefits of powerful systems. Advancing the technology could significantly accelerate medical and scientific research. A unilateral ban could provide China with an enormous advantage in the market. Researcher Gary Marcus also did not support the initiative in its current form. He allowed for a temporary pause and the creation of a separate federal regulator. The scientist described a permanent unilateral ban on superintelligence as too sweeping a step. In August, OpenAI slowed down the development of new neural models. The slowdown was prompted by the growth of cyber capabilities and the Hugging Face incident. The global community is closely watching the progress of this bill.