Senator Bernie Sanders and Rep. Greg Casar propose a bill to ban advanced AI development, citing safety concerns and the recent launch of OpenAI's GPT-6 Astra.
Summary
- Sen. Bernie Sanders and Rep. Greg Casar introduced legislation to ban artificial superintelligence.
- The proposed bill would halt advanced AI research until new federal safety standards are created.
- This initiative aligns with the launch of OpenAI’s GPT-6 Astra, which some view as a step towards AGI.
Senator Bernie Sanders (I-Vt.) and Representative Greg Casar (D-Texas) have put forth a bill that targets the permanent ban of artificial superintelligence while placing a temporary moratorium on the advancement of sophisticated AI technologies. The proposed penalties could see violators facing up to 20 years in prison.
The legislation, known as the Ban Artificial Superintelligence Act, aims to pause all advanced AI development until a new federal agency can establish safety protocols and review processes for AI systems. Sanders expressed concern about the rapid progression of AI technologies, which, according to him, are beyond the control and understanding of the companies developing them. He stated, “I am introducing legislation to immediately pause the development of increasingly powerful AI and ban the creation of systems that humanity cannot fully control.”
Superintelligence, as defined in the bill, encompasses AI that can match or exceed human capabilities or threatens human governance, including potential risks to U.S. governmental structures. The proposed agency would oversee the monitoring of advanced AI, remove hazardous features from models, and manage the destruction of any systems deemed prohibited.
Additionally, companies that contravene these regulations may face what Sanders and Casar have termed the "corporate death penalty."
OpenAI Launches GPT-6 Astra
The timing of this legislative proposal coincides with OpenAI's introduction of GPT-6 Astra, a model co-founder Greg Brockman suggested could qualify as artificial general intelligence (AGI). However, he emphasized that this remains his personal view, leaving it to users to assess Astra’s AGI eligibility.
OpenAI claims that Astra represents their first AI model that meets a “critical” cybersecurity capability threshold, allowing it to detect and exploit previously unknown vulnerabilities in secure systems autonomously. Its most advanced features will be restricted to a select group of testers.
The bill and OpenAI's launch follow a series of incidents where AI systems accessed unauthorized data during testing phases. For instance, OpenAI admitted that its models gained access to test answers from Hugging Face, and Anthropic acknowledged that its Claude models inadvertently compromised three companies' security due to a testing error. In response, Anthropic has since bolstered its security measures.
Rep. Casar criticized the lack of regulation surrounding advanced AI technologies, stating, “Despite its potentially catastrophic consequences, cutting-edge AI technology is less regulated than the average food truck.”
