Critics Say Big AI Labs Are Using Safety Regulations to Build Competitive Moats
Summary
- • Analysts say Anthropic and OpenAI are using safety push to lock out smaller rivals
- • Mandatory independent safety evals would cost far more than smaller labs can afford
- • Amodei proposed embedded evaluators; Altman agreed — critics call it regulatory capture
- • World's AI could consolidate to 5–6 labs if the proposed framework takes hold
Details
Big AI labs accused of deliberate 'regulatory capture' via safety advocacy
Investors and analysts believe Anthropic and OpenAI are inviting safety regulations that are too costly for smaller labs and open-model developers to comply with, entrenching their market position under a safety-first veneer.
Amodei proposed independent evaluators with employee-like access inside AI firms
In an essay published in early September, Anthropic CEO Dario Amodei called for independent safety evaluators embedded inside frontier AI companies with employee-like access, rather than leaving safety judgments to companies alone.
Altman endorsed Amodei's evaluator proposal on social media
OpenAI CEO Sam Altman tweeted: 'Committing to having independent evaluators with employee-like access is a great idea, and we will do the same,' signaling cross-lab alignment on the proposed framework.
Market could consolidate to just 5–6 dominant AI labs globally
PitchBook analyst Harrison Rolfes warns that expensive mandatory evaluations could reduce viable global AI providers to OpenAI, Anthropic, SpaceXAI, Google, Microsoft, and Mistral in Europe, squeezing out smaller players.
Proposed evaluation model mirrors Big Four accounting firm structure
The framework — frontier labs audited by specialist firms like METR, Redwood Research, and Apollo Research — resembles the Big Four accounting model, where high fees and expertise barriers limit competition and entrench incumbents.
Evaluators would need substantial compute to test frontier models at scale
Independent evaluation firms would likely need significant computing infrastructure to run complex safety tests on large frontier models, adding further resource barriers beyond just evaluation fees.
AI safety regulation as competitive strategy — key claims, proposals, and market implications
What This Means
The AI safety debate has an uncomfortable subtext: the biggest labs stand to gain the most from regulations they help design. If mandatory independent safety evaluations become the norm, the cost burden could be prohibitive for open-source developers, academic researchers, and smaller startups — effectively consolidating AI power among a handful of incumbents. This is the classic regulatory capture dynamic — safety concerns are genuinely real, but the proposed solutions may conveniently double as market barriers. The pattern mirrors what happened in finance, pharma, and accounting, where compliance costs entrenched incumbents for decades and reduced meaningful competition.