In a coordinated effort to reshape the regulatory landscape for artificial intelligence, two of the industry’s most prominent players, OpenAI and Anthropic, have ramped up lobbying efforts aimed at securing antitrust protections. According to recently disclosed federal filings, the companies are pushing for the legal right to share information on security risks and adversarial threats, a practice that currently risks running afoul of federal competition laws.
Anthropic’s lobbying strategy has centered specifically on the Collaboration on Adversarial Threats and Security Risks Act. This proposed legislation is designed to allow AI labs to coordinate their responses to security vulnerabilities and external threats without the threat of being accused of market collusion. There is growing momentum for this bill to be folded into must-pass defense policy legislation, reflecting a view within some corners of Washington that AI security is a matter of national security.
OpenAI has taken a broader approach in its recent filings. The company’s disclosures list “AI-security collaboration and information-sharing, including proposed antitrust protections” as primary topics of discussion with federal officials. While lobbying forms do not explicitly state whether a company is supporting or opposing a specific measure, the context of these discussions suggests the labs are seeking a safe harbor to discuss technical safety benchmarks and threat assessments.
The push for antitrust carve-outs comes at a delicate time for the industry. Last month, several AI leaders floated the idea of a coordinated “slowdown” in model development to ensure safety protocols could keep pace with technical advances. This proposal immediately raised red flags for antitrust regulators, as any agreement between competitors to limit production or innovation is typically viewed as a violation of the Sherman Act.
OpenAI’s head of global affairs, Chris Lehane, has recently argued that a specific antitrust exemption for “model pacing”—the intentional slowing of development—is not necessary. However, the idea has already faced sharp criticism from some political circles. Officials in the previous Trump administration argued that AI leaders should be capable of making individual safety decisions without seeking government-sanctioned immunity from competition laws.
Beyond specific antitrust concerns, the lobbying push underscores the massive financial scale of the influence game in the technology sector. While the newer AI labs are making significant inroads, Meta continues to lead the pack in overall Big Tech lobbying expenditures. The industry is currently focused on a wide array of legislative priorities, including export controls on advanced chips and the regulation of massive data centers required to train next-generation models.
As the Collaboration on Adversarial Threats and Security Risks Act moves through the legislative pipeline, the debate will likely center on where the line should be drawn between legitimate security coordination and anticompetitive behavior. For the AI labs, the goal is to create a framework where they can defend against shared threats without the looming threat of federal litigation.
