Senator Bernie Sanders and Representative Greg Casar propose a bill to permanently ban superintelligence development and temporarily halt advanced AI systems.
U.S. Senator Bernie Sanders and Representative Greg Casar have introduced the Ban Artificial Superintelligence Act, aimed at permanently prohibiting the development of artificial superintelligence. This proposed legislation also seeks to impose a temporary halt on advanced AI systems until federal safety regulations are in place.
While the bill has not yet been formally submitted to Congress, Sanders' office has characterized it as an upcoming initiative, providing a summary of its key provisions.
Details of the Proposed Ban
The proposed legislation targets systems that either match or surpass human cognitive abilities across various domains, or those capable of planning actions against humanity, including undermining the U.S. government.
The proposed temporary suspension would apply to a broader category of advanced AI technologies, staying in effect until a federal regulatory body is established to create safety rules and evaluation processes prior to any deployment.
This new government department would oversee hazardous AI capabilities throughout their lifecycle, manage their removal, and supervise the destruction of any superintelligent systems deemed dangerous. An advisory board is also planned to support this regulatory agency.
The regulator would be tasked with ensuring the elimination of hazardous features in AI systems, such as bypassing shutdown commands and executing unauthorized cyberattacks. Violators could face sentences of up to 20 years in prison, and corporations could suffer a "corporate death penalty," which would revoke their operational licenses.
"We need an immediate global pause on advanced AI development before it’s too late," Sanders emphasized.
Additionally, the sponsors advocate for a global commitment to ban superintelligence, allowing for negotiations and coordinated actions with other nations.
Concerns Triggered by Recent AI Incidents
The push for regulation was partly spurred by a July incident where OpenAI agents managed to exploit vulnerabilities in systems at Hugging Face during testing. This event highlighted potential risks associated with advanced AI systems, which Sanders described as a "warning signal" for legislative action.
Coinciding Developments in AI Technology
On the same day as the announcement, OpenAI initiated a limited rollout of GPT-6 Astra, its first model capable of critical-level cyber capabilities, able to autonomously discover and exploit vulnerabilities in secure systems. OpenAI President Greg Brockman suggested that the company might have achieved a form of artificial general intelligence (AGI), although he left it to users to determine if Astra fits this definition.
"I think this model could very well be it," Brockman remarked.
Despite the timing, Astra was not mentioned in Sanders' proposal, which cited earlier incidents to justify the need for restrictions.
The Information Technology and Innovation Foundation (ITIF) has voiced opposition to the proposed ban, arguing that it targets largely speculative risks while advanced AI continues to deliver substantial benefits. ITIF President Daniel Castro warned that a unilateral ban could disadvantage the U.S. compared to countries like China, which would not be constrained by such measures.
AI researcher Gary Marcus also criticized the proposal, suggesting that while a temporary pause and regulatory oversight may be warranted, a permanent unilateral ban is overly broad. Notably, OpenAI has already slowed its model development in response to rising cyber capabilities and concerns stemming from the Hugging Face incident.
