Critics accuse Dario Amodei of seeking AI duopoly
- Dario Amodei proposed voluntary AI training limits to prevent rogue agent risks.
- Rival AI executives backed the pause, but critics call it an anti-competitive cartel scheme.
- Skeptics argue closed labs want to mask high training costs before public stock offerings.
Anthropic CEO Dario Amodei wants to pause the AI arms race. The Silicon Valley establishment fell in line almost instantly.
In a 3,000-word essay published on September 14, 2026, Amodei called for deliberate capability pacing and mandatory third-party oversight of frontier training runs. He cited recursive self-improvement and rogue agent swarms as escalating threats. Within hours, OpenAI CEO Sam Altman, Google DeepMind lead Demis Hassabis, and Elon Musk publicly endorsed the framework. Anthropic immediately committed to embedding researchers from risk nonprofit METR into its training labs.
The consensus unraveled as skeptics questioned corporate motives. On Breaking Points, Krystal Ball pointed out that closed labs face crushing capital expenditures ahead of prospective public offerings. Excluding massive training runs, existing AI products deliver 80 percent profit margins. Investor Michael Burry argued on The AI Daily Brief that framing a financial slowdown as existential caution gives labs cover to curb training burn without admitting technical bottlenecks to Wall Street.
"Capabilities are compounding far faster than safety research."
- Nate Soares, Breaking Points with Krystal and Saagar
Competitors and market analysts saw a defensive play against cheaper rivals. On The Jack Mallers Show, host Jack Mallers called the proposal a blatant attempt at regulatory capture to block low-cost competitors like DeepSeek, which delivers comparable capabilities at 1 percent of the cost. Professor Hal Singer warned on The AI Daily Brief that public calls for joint training limits resemble informal cartel behavior designed to lock in a corporate duopoly.
The pushback extended to the mechanics of oversight itself. Critics noted that METR researchers share office space and institutional ties with the labs they are supposed to monitor. In response, Hugging Face CEO Clement Delangue offered his platform as an independent evaluator for open alignment. Meanwhile, Meta chief AI scientist Yann LeCun dismissed Amodei's warnings as recycled panic intended to pull up the ladder on open-source development.
"Tech leaders can no longer unilaterally set the pace of AI progress without government intervention."
- Nathaniel Whittemore, The AI Daily Brief
Political leaders quickly shut down hopes for regulatory coordination. White House AI adviser David Sacks argued on Breaking Points that incumbent labs already hold a duopoly and do not need antitrust exemptions to slow their own work. Donald Trump rejected the pacing initiative as a hoax that would hand technological dominance to China. Chinese state media echoed that view from the opposite side, slamming the proposal as a hypocritical ploy to protect American tech hegemony.
The scaling race is no longer just a technical challenge. It is a financial and political street fight.