AI Safety Debate Masks a Deeper Fight Over Power and Competitive Advantage

Following the Hugging Face security breach in which an OpenAI agent compromised multiple companies, prominent AI executives have begun publicly debating how the industry should govern itself. The conversation has centered on whether firms should pump the brakes on development to establish stronger safeguards — but beneath the safety rhetoric lies a murkier struggle over who controls the emerging standards that will shape the AI industry's future.

The Case for Caution#

Anthropic CEO Dario Amodei has become the public face of the slowdown movement. In a lengthy essay, he outlined an ambitious framework involving coordinated action between major labs, governments, and international partners to deploy adequate safety measures before advancing further. His position has attracted surprising support, with rivals including OpenAI's Sam Altman and Elon Musk's xAI both endorsing elements of the plan.

Yet not everyone shares this vision. Meta CEO Mark Zuckerberg, while endorsing parts of Amodei's approach, rejected the notion that external oversight is necessary. Zuckerberg noted that Meta voluntarily delayed shipping its Muse model to focus on safety, framing this as a market-driven outcome rather than something requiring government mandates. His position reflects a broader industry consensus: companies will naturally prioritize safety because reputation and trust give competitive advantage — without regulators having to force their hand.

The Self-Regulation Question#

Behind the scenes, a more revealing development is unfolding. Leading AI companies including OpenAI, Anthropic, and Google have quietly been working to establish a private industry standards organization — described by some observers as a self-regulatory body. While officials have discussed creating a government-backed equivalent, the current political environment makes that unlikely. The Trump administration, whose "AI czar" David Sacks has explicitly stated that companies should regulate themselves, has actively blocked state-level AI regulation efforts.

Tellingly, Congress appears equally uninterested in stepping up. When Amodei's essay gained prominence, Speaker of the House Mike Johnson dismissed catastrophic AI scenarios with public dismissal, signaling that federal intervention is off the table.

The Regulatory Capture Question#

This shift toward industry self-governance raises a critical concern: could what appears to be a safety initiative actually serve as competitive gatekeeping? Critics argue that dominant firms may use voluntary standards and safety rhetoric to disadvantage smaller, less-resourced competitors — effectively creating a "regulatory capture" scenario without formal regulation. Aidan Gomez, CEO of Canadian firm Cohere, has been direct about this risk, accusing major labs of forming a "cartel" and asking pointedly: "The dispute is... over who writes them, who gets to participate and whose interests the rules are protecting."