Comment on FTC/DOJ Antitrust Guidelines for Collaborations Among Competitors

Box Commons urges the Department of Justice and Federal Trade Commission to provide clear antitrust safe harbors for AI credentialing standards development organizations. The 2000 Antitrust Guidelines for Collaborations Among Competitors were withdrawn in 2023, leaving AI standards bodies without the analytical framework they need to evaluate whether their credentialing activities constitute lawful standard-setting or risk being characterized as unlawful group boycotts.

This gap is not hypothetical. AI agent credentialing requires competitor collaboration — technology companies, insurers, and patent holders working together in an SDO to establish behavioral safety standards. The antitrust exposure of such collaboration is currently indeterminate.

The AI ecosystem currently has no widely adopted credentialing standard for behavioral safety. The result is a market failure: deployers cannot verify the safety of AI systems they procure, insurers cannot price risk consistently, and regulators lack evaluation infrastructure. Standards bodies exist to solve exactly this kind of coordination problem — but only if antitrust law permits the necessary competitor collaboration.

When an SDO denies a credential to an AI system that fails its behavioral safety standards, the denied party may characterize the denial as a group boycott. Without clear safe harbor guidance, SDOs face the choice between watering down standards to avoid antitrust exposure and maintaining rigorous standards at legal risk.

We propose four specific safe harbors:

1. SDO Safe Harbor for AI Security Standards. Standards development activities that follow consensus procedures (public comment, balanced representation, appeals mechanisms) should receive per se legality, consistent with the Standards Development Organization Advancement Act of 2004.

2. Collective AI Threat Intelligence Sharing. Sharing of behavioral safety threat data (failure modes, adversarial attack patterns, safety-critical incidents) among competitors should be treated as procompetitive joint venture activity.

3. Credentialing Decisions Meeting Procedural Safeguards. Credential denials by SDOs that maintain published criteria, independent assessment, and formal appeals processes should be evaluated under the rule of reason, not treated as per se group boycotts.

4. Insurance-Linked Credentialing. Credentialing activities that support insurance underwriting — enabling insurers to price AI risk based on verified behavioral safety rather than blanket exclusions — should be recognized as procompetitive.

Large technology companies can absorb the legal costs of antitrust uncertainty. Small, mission-driven standards bodies cannot. Without clear safe harbors, antitrust ambiguity functions as a barrier to entry in the standards development market — concentrating standard-setting authority in organizations with the legal resources to navigate uncertainty, regardless of whether those organizations produce the most rigorous or most independent standards.

The Commission should recognize that antitrust clarity for AI credentialing is itself procompetitive — it enables a more diverse and independent standards ecosystem.