OpenAI Asks Congress Whether an AI Slowdown Would Be Legal
OpenAI has asked U.S. lawmakers whether rival AI developers could lawfully coordinate a slowdown in model development, after its chief scientist called for voluntary pauses until safety can be better demonstrated. The request comes amid debate over antitrust risks, lawmakers' proposals to carve out protected security collaborations, and warnings that competitive and financial pressures push companies to move quickly despite safety concerns.
Why It Matters
The question intersects antitrust law, national security and AI safety: clarifying whether companies can legally coordinate pauses would shape whether industry-wide slowdowns are possible as a tool to manage systemic AI risks. At the same time, experts warn that market and geopolitical competition incentivize firms to press ahead, potentially undermining unilateral restraint.
Key Facts
- OpenAI outreach: OpenAI asked members of Congress in recent weeks whether an agreement among AI companies to slow development could violate antitrust laws.
- Safety appeal from OpenAI: OpenAI chief scientist Jakub Pachocki urged the industry to adopt shared safety standards and called for voluntary slowdowns until systems can be shown safer.
- Antitrust and legislation: Senators Adam Schiff and Jim Banks introduced a bill that would protect certain security collaborations to restrict AI development, subject to advance notice to the Justice Department.
- Expert warnings on competition: Miranda Bogen of the Center for Democracy and Technology said commercial and geopolitical competition creates incentives to release products before risks are understood.
- Industry pressure and funding: Duncan Sabien of the Machine Intelligence Research Institute said each advance brings substantial funding and influence, creating pressure to continue developing rather than unilaterally slowing.
OpenAI recently queried U.S. lawmakers about whether rival AI developers could legally agree to slow down model development, raising antitrust questions as the company’s chief scientist urged industry-wide restraint until safety can be better demonstrated. The outreach follows public calls from OpenAI scientists for shared safety standards and comes amid growing discussion about how to manage systemic risks from increasingly capable systems.
Legal experts and lawmakers are weighing the tension between antitrust law and coordinated safety efforts. An antitrust concern is that an agreement to limit development could suppress competition; in response, Senators Adam Schiff and Jim Banks have proposed legislation that would shield certain security collaborations from liability if participants notify the Justice Department in advance. OpenAI’s inquiry to Congress appears aimed at clarifying whether coordinated slowdowns would face legal exposure under current rules.
Researchers and policy specialists say market and geopolitical forces make voluntary pauses difficult to sustain. Miranda Bogen of the Center for Democracy and Technology told reporters that intense commercial and geopolitical rivalry pushes companies to release products before fully understanding their risks. Similarly, Duncan Sabien of the Machine Intelligence Research Institute argued that each technical advance attracts major funding and expands a developer’s influence, which discourages unilateral restraint when rivals may continue advancing.
The debate has unfolded alongside other industry developments. Both OpenAI and Anthropic eased some safety commitments in February, while OpenAI paused internal work on a project called Astra in August over cybersecurity concerns. The broader conversation escalated this week after a former Anthropic engineer publicly resigned, saying he feared continued AI development could pose catastrophic risks within the decade. Policymakers, researchers and companies now face the question of whether legal clarity or new regulatory carve-outs are needed to enable coordinated safety measures without running afoul of competition law.
Keep Reading
An Anthropic researcher’s doomsday warning comes at a very interesting time
Claude users found ways around safeguards for bioweapons research
Where to preorder the Apple AirPods 5
