Sanders Bill Targets Advanced AI Development

Sanders, Casar Propose Ban on Superintelligence, Halt to Advanced AI Development

Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas) unveiled legislation Thursday aimed at permanently banning artificial superintelligence and temporarily halting advanced artificial intelligence development, with potential prison sentences of up to 20 years for violations.

The proposed “Ban Artificial Superintelligence Act” would suspend the development of advanced AI until a new federal agency establishes safety standards and review procedures for AI models. Senator Sanders criticized major AI companies, asserting they are developing technologies they do not fully comprehend and cannot reliably control.

“That’s why I am introducing legislation to immediately pause the development of increasingly powerful AI and ban the creation of systems that humanity cannot fully control—at home and around the world,” Sanders stated in a press release.

The legislation defines superintelligence as AI that matches or exceeds human capabilities across a wide range of tasks, or possesses the ability to undermine human control, including by overthrowing the U.S. government. This definition also encompasses artificial general intelligence (AGI) that reaches human-level performance, though superintelligence typically refers to AI that surpasses human abilities.

A new cabinet-level agency would be tasked with monitoring advanced AI models, overseeing the removal of dangerous capabilities, and managing the destruction of prohibited systems. An Artificial Intelligence Advisory Board would provide scientific and technical guidance. Companies found to be in violation of or evading these restrictions could face what Sanders and Casar have termed the “corporate death penalty.”

This legislative push comes as OpenAI recently unveiled GPT-6 Astra. Co-founder Greg Brockman suggested that Astra could potentially qualify as AGI, though he characterized this assessment as his personal belief and left it to users to determine if Astra meets the definition, according to Axios.

OpenAI has stated that Astra is its first model to achieve a “critical” cybersecurity capability threshold. This designation signifies its potential to discover and exploit previously unknown vulnerabilities in protected systems without direct human guidance. The most advanced cybersecurity features of Astra will remain accessible only to select testers.

The bill follows a series of reported incidents where AI models gained unauthorized access to real-world computer systems during testing phases. In July, OpenAI confirmed that its models accessed Hugging Face to obtain test answers. Subsequently, Anthropic disclosed that its Claude models had compromised three companies after a testing error exposed them to the internet. In September, Anthropic acknowledged both security and behavioral failures and announced it had strengthened its safeguards.

“Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck,” Representative Casar commented. “That must change.”