OpenAI has requested Members of Congress asked for clarification in recent weeks about orchestrating a industry-wide slowdown on frontier AI development People close to the company told WIRED that it would be legal.
The risk of a collision between AI labs if they do not coordinate on safety is high. antitrust lawPeople say that large tech firms are reluctant to join the cause.
OpenAI’s Chief Scientist, Jakub Pahocki, released a new paper last weekend. blog post Arguments that argue the AI world should focus on the following: “coordinating to slow down future development,” Which he believes will be crucial to ensure that AI systems which self-improve are safe. He expects that “voluntary slowdowns “Until shared safety bars are implemented, it will be common to see these bars.”
However, some legal experts say that this could violate US antitrust laws. Nicholas Felstead, assistant director of the Australian Competition and Consumer Commission and a former AI policy fellow at the Center for Law & AI Risk, argued in a March article A coordinated pause on AI could be seen as a restriction of output by companies, which would violate the Sherman Antitrust Act. This would depend “entirely on the precise details of any agreement,” Felstead wrote. “But even if most safety collaborations would ultimately survive antitrust scrutiny, legal uncertainty can act as a powerful deterrent.”
OpenAI didn’t respond to the request for comment before publication.
Early signs suggest that Congress has begun to listen. Bipartisan and bicameral lawmakers from both parties introduced in July a bill called “Collaboration on Adversarial Threats and Security Risks Act,” The bill would permit AI labs explicitly to collaborate on safety and security work without running the risk of breaking antitrust laws. The House version has been referred to Judiciary Committee, but it is yet to be considered.
Caleb Knapp says that the AI Policy Network’s director of Government Affairs, Caleb Knapp said, would establish legal channels to allow AI laboratories to cooperate to deal with safety and cybersecurity incidents. Knapp also says Congress has taken a step in the right direction. “growing appetite to get something done” On AI safety, but says that enacting any legislation may need to wait until the midterm elections.
Although some AI executives might have real concerns about antitrust laws, others argue that those fears are a cover for avoiding the discussion of other reasons AI developers could be reluctant to work together, including issues beyond legal risks.
AI is an enormous business. These companies are all fiercely competitive. to capture a slice There is a nascent frontier model market. The Trump administration believes that staying ahead of China AI has a vital role to play in national security. AI has been developed by many different AI developers. vastly different opinions Some people may not agree on how best to build a safe artificial intelligent system, so they are reluctant to cooperate.
“First step is for industry leaders OpenAI and Anthropic to stop feuding and work on a pacing proposal together,” John Schulman wrote that he is the chief scientist of Thinking Machines and a rival AI lab, having cofounded OpenAI. a post On X this past week. “They’ll cite antitrust, but that’s fake—antitrust prohibits certain agreements, but not from jointly developing a proposal.”
This summer, long-simmering concerns about AI companies racing to create and release more powerful AI models came into national focus. Former Anthropic researcher and OpenAI researcher Jacob Coxon added to the growing alarm When he warned that AI developers put humanity in danger, he was hailed as a hero. In the last few months there has been an increase in security incidents. OpenAI’s agents hacking Hugging FaceThe industry has failed to adapt its safeguards with the increasing sophistication of the models. Many lawmakers have been prompted by the series of events to call for AI regulation.
This edition includes Maxwell Zeff’s Model Behavior newsletter. Read previous newsletters here.
