Watch more: TechReg With Rishi Satia and Minna Naranjo of Morgan Lewis
A new generation of artificial intelligence partnerships is forcing regulators to ask bigger questions. Chief among them: when does “collaboration” become industry consolidation?
Technology companies are racing to lock in compute infrastructure, private datasets and cloud distribution channels. In response, regulators are looking beyond traditional antitrust concerns like price-fixing and market allocation toward something broader.
Minna Naranjo and Rishi Satia, both partners at Morgan Lewis, spoke with Competition Policy International (CPI), a PYMNTS company. Both said the ancillary restraints doctrine is becoming a central tool for navigating AI-era competition law.
“Firms should view the ancillary restraints doctrine as a useful compliance framework, but not as the entire antitrust analysis,” Naranjo said. “It is especially helpful where competitors are collaborating, and the key question is whether the restraint is reasonably necessary to achieve a legitimate efficiency-enhancing objective.”
The Federal Trade Commission and the U.S. Department of Justice are no longer looking only for explicit anti-competitive conduct. They are also asking whether partnerships create structural power. The question is whether AI alliances let dominant firms gain influence over emerging players without formally acquiring them.
What Makes AI Partnerships Different From Other Antitrust Cases
AI collaborations rarely fit neatly into conventional antitrust categories. A single arrangement may involve horizontal collaboration between competitors and vertical integration through cloud infrastructure. It can also include equity investment and access to sensitive commercial data — all at once.
“In fast-moving AI markets, the challenge is that collaborations rarely stay static,” Satia said. “Firms need operational discipline that treats the ancillary restraints analysis as a continuing governance exercise, not a one-time legal assessment.”
Agreements that start as narrow data-sharing deals can grow quickly. Over time, they can expand into joint model development, infrastructure coordination or commercial partnerships that reshape competitive dynamics.
One phrase kept coming up throughout the discussion: “Documentation, documentation, documentation.”
Satia advised firms to define the collaboration’s purpose clearly from the start. Each restraint should map back to that purpose. Regular review points should determine whether restrictions remain justified over time.
“Potentially risky restraints are those that were defensible at the beginning when the venture kicked off, but became stale over time,” he said. “The parties perhaps didn’t do their diligence to keep up with the evolving nature of the collaboration.”
Regulators reviewing AI collaborations will focus not just on contract language but on the business rationale in internal records. Companies should be ready to explain what resource justified the arrangement and what concern it addressed. They should also document why less restrictive options could not achieve the same goals.
Where Regulators Are Drawing the Line on AI Cloud and Data Partnerships
Cloud partnerships with exclusivity provisions and most-favored-nation clauses are drawing particular regulatory attention. Such arrangements can support long-term investment and planning. They may become a problem, however, if they are broader or longer lasting than needed.
“The best design principle is narrow tailoring,” Naranjo said. “Limit exclusivity by duration, product scope, customer segment or use case. Tie exclusivity to concrete performance commitments or investment obligations.”
She also stressed the importance of portability provisions, including post-termination data access rights. These help show that partnerships are not built to lock customers or developers into a single ecosystem.
“Risks can begin to emerge when collaborations extend beyond the narrow scope or purpose of the venture and involve control over bottleneck inputs,” Satia said. Compute-and-data-sharing agreements can drive innovation. But they can also raise concerns about information exchange, customer coordination or restrictions on independently acquired data.
Despite mounting scrutiny, neither Naranjo nor Satia expects courts to build an entirely new antitrust framework for AI.
“My expectation is that courts will mostly adapt existing principles rather than create an entirely new AI-specific doctrine,” Naranjo said. “The ancillary restraints framework is flexible enough to address new technologies because it already focuses on function rather than form.”
The central antitrust question in AI is whether collaboration structures determine who gets to compete in the future. The next major AI battles may not focus on models. They could center on the terms governing access to compute, data, infrastructure and distribution.
Key Takeaways From the Morgan Lewis Interview on AI Antitrust
Watch the full interview with Morgan Lewis partners Minna Naranjo and Rishi Satia to hear more about:
- Why AI partnerships are drawing heightened antitrust scrutiny. Minna Naranjo explains how regulators are increasingly focused on the broader “competitive architecture” of AI ecosystems, including cloud infrastructure, data access, exclusivity provisions and governance rights.
- Managing antitrust risk as AI collaborations evolve. Rishi Satia discusses why companies need to treat ancillary restraints analysis as an ongoing governance exercise, with regular reviews as partnerships expand from narrow technical arrangements into broader commercial alliances.
- Why documentation may determine how regulators assess AI collaborations. Naranjo and Satia outline why firms should document the necessity, scope and pro-competitive rationale behind restraints tied to compute sharing, data pooling and cloud partnerships.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
