What happened
According to WIRED, OpenAI is investigating the legal feasibility of coordinating with other AI labs to slow down development, a strategy proposed by its chief scientist to ensure safety. While a bipartisan bill introduced in July aims to provide antitrust immunity for such collaborations, it has not yet been enacted, leaving the industry in a state of legal uncertainty.
WIRED reports that OpenAI is currently evaluating whether a coordinated slowdown in AI development would be legal under US antitrust laws. This inquiry follows a blog post by OpenAI chief scientist Jakub Pachocki, who argued that coordinating to slow down future development is essential for ensuring the safety of self-improving AI systems. Pachocki suggested that voluntary slowdowns should become commonplace until shared safety standards are established.
Legal experts cited by WIRED, including Nicholas Felstead, note that such coordination could potentially violate the Sherman Antitrust Act by restricting output. While the legality would depend on the specific details of any agreement, the resulting legal uncertainty acts as a powerful deterrent for companies considering safety collaborations. OpenAI did not respond to WIRED's request for comment on this matter.
In response to these concerns, a bipartisan group of lawmakers introduced the 'Collaboration on Adversarial Threats and Security Risks Act' in July. This bill would explicitly permit AI labs to coordinate on security and safety work without facing antitrust risks. However, the House version of the bill has been referred to the Judiciary Committee and has not yet been taken up for further action.
Caleb Knapp of the AI Policy Network, which endorsed the bill, stated that Congress has a growing appetite for AI safety legislation but that enactment may be delayed until after the upcoming midterm elections. Meanwhile, some AI leaders, including John Schulman, argue that antitrust concerns are a convenient excuse for companies to avoid addressing deeper competitive and philosophical disagreements about AI safety.
Why it matters
This legal ambiguity creates a significant obstacle to collective action on AI safety. Without clear legal protection, major tech companies may avoid coordinating on safety measures due to the risk of antitrust lawsuits, potentially allowing the competitive race for more powerful models to continue unchecked despite growing concerns about security and societal risks.
The potential for antitrust liability creates a significant barrier to industry-wide safety coordination. As AI capabilities grow, the risk of security incidents and societal harm increases, yet the legal framework currently discourages the very collaboration needed to mitigate these risks.
This legal uncertainty may allow the competitive race for frontier models to continue without meaningful safety checks. Companies may prioritize speed and market share over collective safety efforts, particularly if they believe that coordinating with rivals could expose them to legal challenges.
The situation highlights a critical gap in current US policy. While there is bipartisan interest in AI safety, the lack of enacted legislation leaves AI labs in a precarious position, unable to confidently engage in the kind of coordination that experts argue is necessary for safe development.
What to watch next
Monitor the progress of the 'Collaboration on Adversarial Threats and Security Risks Act' in the House Judiciary Committee and watch for any public statements from OpenAI or other major labs regarding their willingness to engage in safety coordination once legal clarity is established.
The next steps for the 'Collaboration on Adversarial Threats and Security Risks Act' in the House Judiciary Committee will be a key indicator of congressional commitment to resolving the legal ambiguity surrounding AI safety coordination.
Public statements from OpenAI and other major AI labs regarding their stance on safety coordination will be closely watched, particularly as the legal landscape evolves and the competitive pressure to release new models intensifies.
Any further security incidents involving AI models, such as the recent Hugging Face breach, may increase pressure on lawmakers to act quickly and provide clear legal guidance for the industry.