OpenAI supports a set of bills aimed at preventing the rise of threats from artificial intelligence in the realm of biological weapons and synthetic viruses. These bills consider improving access to essential data for building AI defenses against both natural and artificial biological threats and establish standards for the safe development of these tools.
Details of the Bills and Their Objectives
OpenAI has recently supported three bipartisan bills: the "Biological Data Network Act," the "AI-Ready Biological Data Standards Act," and the "SCALE Biology Act." The first two proposals have already been introduced in the House of Representatives and the Senate, while the last one has only been introduced in the House. OpenAI's support for these bills reflects the company's expanding efforts to legislate in areas related to biological security, especially with growing public concerns about the destructive potential of advanced AI models.
Risks and Challenges in Using AI
Recent research has drawn attention to the use of AI for designing pathogens. For example, a notable study by scientists at Stanford University and the Arc Research Institute showed that they were able to design 16 new viruses using this technology. Additionally, Anthropic revealed last week five instances of illegal use of its models for pathogenic research that could aid in the development of biological weapons. Two of the bills that OpenAI supports expand the oversight of the National Institute of Standards and Technology (NIST) in matters related to AI and biological security.
The SCALE Biology Act establishes a new program at NIST tasked with creating measurement standards for biological data and guidelines for the safe development of biotechnologies and preventing their misuse. The AI-Ready Bio-Data Standards Act also mandates NIST to design a framework for standards to ensure that government-sourced biological data is accessible to AI researchers.
The Biological Data Network Act also requires the Department of Energy to create a centralized source for accessing biological data. Richard Johnson, OpenAI's National Security Risk Reduction Officer, emphasized the importance of the data and standards outlined in these bills for developing better assessment methods, stating, "We need to be able to continuously measure capabilities and risks and then compare where we are at any given time."
Kaitlin Fraser, Executive Director of the National Security Commission on Emerging Biotechnologies, also stated, "High-quality, AI-ready biological data should be regarded as a strategic national resource." OpenAI has previously supported two other related bills named the "Biological Security Innovation and Modernization Act" and the "Homestake AI Act."
Potential biological threats arising from AI have intensified calls from the industry to slow the advancement of this technology. Johnson has stated that such laws help AI research better understand what the pace of development means for biological security. He said, "We need to ensure we understand the capabilities and make sure we are not moving so fast that we outpace our ability to understand how to use these things and their impact on biological security."
