Imagine three competing CEOs worried their product could kill people, joining a conference call to discuss pausing production — and then lawyers step in. That scenario actually occurred in March 2020, when the chief executives of General Motors, Ford, and Fiat Chrysler tried to coordinate production halts amid Covid–19 concerns; their lawyers warned that such coordination might violate antitrust laws. Although legal objections were later set aside and Detroit factories shut down, the episode highlights the tension now playing out in the AI sector.
The present tension: a proposed slowdown and antitrust law
The current dispute centers on Dario Amodei's request for a narrowly tailored antitrust waiver that would permit coordinated slowdowns in the development of AI models. The proposal has sparked debate about whether the industry is invoking existential-risk rhetoric to justify closing ranks and protecting market position.
Those skeptical of that line argue it is cynical: executives did not suddenly discover existential risk last week to bolster IPO or strategic plans. Many industry leaders have been discussing AI safety for years, and their concerns may be longstanding and genuine.
Why an "AI OPEC" is unlikely
Several factors make a formal, cartel-like AI grouping improbable:
- Mark Zuckerberg has publicly rejected participation in such an arrangement, indicating Meta would not join.
- Intense competitive culture among tech executives and entrenched market rivalry make durable collusion difficult, even for firms that are already wealthy enough not to need to win.
- Any coordination that does emerge is more likely to resemble patterns in the airline industry: as a base product commoditizes, companies compete on reliability and bundled services. Notably, airlines also compete on safety because accidents are detrimental to all players.
Even the real OPEC has limitations: it failed to prevent the United Arab Emirates walking away and does not include three of the world's four largest oil producers.
AI may be pro-competitive rather than monopolistic
Contrary to fears of market enclosure, AI is increasingly seen as lowering the cost of starting many kinds of businesses, which can have pro-competitive effects. This perspective suggests that an overly restrictive application of antitrust rules might block useful cooperation without necessarily improving safety.
Diverging corporate views and public criticism
The debate also includes sharp disagreements among firms. Reuters reported that Mustafa Suleyman, co‑head of Microsoft’s AI efforts, criticized Anthropic’s approach to training its Claude model on ideas related to consciousness. Such criticisms show that companies contest both the means and the ends of safety approaches in public, rather than forming a single united front.
Conclusion
Legal restraints and market incentives will strongly influence how industry discussions about AI safety proceed. Historical examples, statements from major executives, and market dynamics together indicate that a formal "AI OPEC" is unlikely to materialize, but antitrust law and competition policy must be calibrated so they do not inadvertently block meaningful, timely safety collaboration.
