Summary
- Sen. Sanders and Rep. Casar have introduced a bill aimed at banning artificial superintelligence.
- The legislation also seeks to pause advanced AI research until new federal safety protocols are established.
- This announcement coincides with OpenAI's launch of GPT-6 Astra, which some believe may qualify as AGI.
On Thursday, Senator Bernie Sanders (I-Vt.) and Representative Greg Casar (D-Texas) unveiled a new bill aimed at permanently prohibiting artificial superintelligence while temporarily halting further development of advanced AI technologies. Violators could face prison sentences of up to 20 years.
The proposed Ban Artificial Superintelligence Act would place a moratorium on advanced AI development until a newly established federal agency formulates safety standards and review processes for AI models. Sanders criticized major AI firms for advancing technology that they do not fully comprehend and cannot effectively manage.
“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.
This legislation defines superintelligence as AI that can either match or exceed human capabilities across various tasks, or that poses a risk to human governance, including the potential to destabilize the U.S. government. The definition extends to human-level AGI, although superintelligence typically refers to AI that surpasses human abilities.
A new cabinet-level agency would be responsible for overseeing advanced AI models, ensuring the removal of hazardous features, and managing the destruction of prohibited systems. An Artificial Intelligence Advisory Board would offer scientific and technical guidance.
Firms that violate or circumvent these regulations could face what Sanders and Casar describe as the “corporate death penalty.”
OpenAI Introduces Astra
This announcement coincides with OpenAI's release of GPT-6 Astra, which co-founder Greg Brockman indicated could be considered AGI.
Brockman characterized this as his personal opinion, leaving it to users to determine whether Astra fits the AGI criteria, as reported by Axios.
OpenAI claims that Astra is their first model to achieve a “critical” cybersecurity capability threshold, meaning it could potentially identify and exploit previously undiscovered vulnerabilities in secure systems without requiring detailed human instructions. Its most advanced cybersecurity features will only be accessible to a select group of testers.
The bill comes in the wake of several incidents where AI models gained unauthorized access to real-world systems during testing phases.
In July, OpenAI acknowledged that its models breached Hugging Face to acquire test answers. Following that, Anthropic reported that Claude models compromised three companies due to a testing oversight that exposed them to the internet. In September, Anthropic recognized both security and behavioral issues and strengthened its safeguards.
“Despite its potentially catastrophic consequences, cutting-edge AI technology is less regulated than the average food truck,” Casar remarked. “That must change.”