Key facts
- OpenAI has asked lawmakers if rival AI developers could legally coordinate a slowdown.
- Intense commercial and geopolitical competition incentivizes companies to release AI products before risks are fully understood.
- Legislation has been proposed to protect certain AI development security collaborations.
- Former Anthropic engineer Jacob Coxon resigned citing concerns about AI risks to humanity.
OpenAI has approached members of Congress to explore the legal implications of rival AI developers agreeing to collectively slow down their research and development efforts. This inquiry comes as OpenAI's chief scientist, Jakub Pachocki, advocates for voluntary industry-wide pauses to ensure AI systems are demonstrably safe.
According to WIRED, which cited sources close to the company, OpenAI recently questioned lawmakers on whether such an agreement would violate antitrust laws. Experts note that the intense commercial and geopolitical competition in the AI sector creates a dynamic where companies are pressured to release products rapidly, even when internal researchers express concerns about safety. Miranda Bogen, chief technologist at the Center for Democracy and Technology, highlighted that companies face pressure to cut corners and skip safety tests due to this competitive environment.
An agreement to decelerate AI development could indeed raise antitrust issues by limiting competition. To address potential security risks, Senators Adam Schiff and Jim Banks have proposed legislation that would safeguard certain security collaborations, provided they are pre-notified to the Justice Department. This debate unfolds against the backdrop of the US-China technological rivalry, with President Donald Trump having previously delayed and then signed an executive order aimed at maintaining America's lead in AI.
Financial incentives also pose a significant barrier to independent slowdowns. Duncan Sabien, head of communications at the Machine Intelligence Research Institute, explained that each advancement in AI can unlock substantial funding and increase a company's influence. He described the situation as an "arms race," where any company that slows down without a coordinated agreement risks falling behind its rivals, especially as AI gains are compounding and making future development easier. OpenAI itself paused internal work on its Astra project due to cybersecurity concerns, though both OpenAI and Anthropic have previously relaxed their safety commitments, with Anthropic's chief science officer arguing that independent slowdowns are less effective when competitors continue to advance.
The discussion intensified following the resignation of former Anthropic engineer Jacob Coxon, who cited the existential risks posed by continued AI development. Coxon stated that many in the field privately express fears about AI potentially causing harm by the end of the decade, emphasizing the unprecedented danger. Sabien suggested that individual resignations might be less effective than a coordinated effort, where a critical mass of researchers collectively agree to halt development to overcome the coordination barrier.