The news
Cohere CEO Aidan Gomez has rejected calls for the largest frontier AI labs to coordinate on safety standards. He described the idea as a potential cartel that would lock in the advantages of companies already ahead in compute and funding. Gomez made the comments during an interview on Bloomberg Tech with Ed Ludlow on September 14.
Context
Proposals for shared safety rules among top labs have surfaced in recent months as governments and researchers press for tighter controls on advanced models. Gomez accepts that stronger safeguards, independent testing, and greater transparency are necessary. He argues the process cannot be limited to a handful of firms such as Anthropic and OpenAI if the outcome is to remain credible and open to competition.
The single-source interview does not detail any specific proposal under discussion or name additional companies that might join such an effort. It focuses instead on the structural risk Gomez sees in letting the biggest players write the rules.
Details
Gomez stated that tighter rules are required but must come from a broader group of stakeholders. He warned that limiting the conversation to the largest labs would entrench their market power and amount to a “cartel.” The remarks were delivered on the Bloomberg Tech program.
No numbers on model sizes, compute budgets, or timelines appear in the available summary of the interview. Gomez did not outline an alternative governance structure or list which additional organizations should participate. The core claim rests on the distinction between necessary technical safeguards and the process by which those safeguards are set.
Why it matters
Gomez’s position draws a direct line between safety governance and market concentration. When standards are written by the firms with the largest compute budgets, compliance costs rise for everyone else. Smaller labs and startups must then meet requirements shaped around the architectures and resources of the leaders, without equivalent input or scale. This dynamic favors incumbents regardless of whether the rules themselves are technically sound.
The interview frames safety coordination as a question of who holds power rather than a purely technical exercise. If frontier labs align on testing protocols or disclosure norms without wider participation, the result can function as a barrier even when presented as neutral best practice. Regulators and standards bodies will hear competing arguments: one side will claim collective action by the largest players improves safety for all; the other will claim it simply cements existing advantages in funding, talent, and infrastructure.
Cohere’s stance does not reject oversight. It rejects oversight designed in a closed room. That distinction matters for companies that must decide whether to support industry-led initiatives or push for government or multi-stakeholder processes instead. The choice will shape which organizations can still ship frontier models five years from now and which will face compliance costs they cannot meet.
---
Sources:
{
"publisher": "Bloomberg Technology",
"title": "Cohere CEO Warns Against an AI Safety ‘Cartel’",
"url": "https://www.bloomberg.com/news/videos/2026-09-14/cohere-ceo-warns-against-an-ai-safety-cartel-video",
"published_at": "2026-09-14T17:55:52.000Z",
"summary": "Cohere CEO Aidan Gomez is pushing back on proposals for the biggest frontier AI labs to coordinate on safety standards, warning that such an approach could entrench dominant players and amount to a “cartel.” Gomez argues that tougher safeguards, independent testing and greater transparency are needed, but says the rules should be shaped by a broader group of stakeholders rather than companies such as Anthropic and OpenAI alone. He joins Ed Ludlow on \"Bloomberg Tech.\" (Source: Bloomberg)"
}
No comments yet