AI Pause Would Help Dominant Firms, Not Safety, Think Tank Warns

The libertarian Cato Institute cautioned that a government-enforced halt on AI development would likely protect incumbent firms and slow deployment of beneficial technologies. In a blog post, Cato technology policy scholar Jennifer Huddleston said companies can mitigate specific hazards through voluntary measures and common standards instead of an across-the-board pause. She warned that regulatory regimes written for current models may be slow to adapt and could obstruct future improvements that might be safer.

Huddleston also argued that rules shaped around the capabilities and resources of today’s leading developers could consolidate their market positions, making it harder for smaller firms to compete. She raised national-security concerns as well, suggesting that a U.S. pause could weaken cyberdefenses while foreign competitors continue advancing. Those comments came amid partisan and industry debates: lawmakers including Sen. Bernie Sanders and Rep. Greg Casar have proposed a development pause and an outright ban on artificial superintelligence, while OpenAI CEO Sam Altman has urged a slower, precautionary approach and stronger safeguards ahead of federal regulations.

Block co-founder and chairman Jack Dorsey expressed similar worries about entrenching dominant firms, arguing on X that safety-focused rules written by today’s market leaders could become barriers to entry. Dorsey said he favors open releases that researchers can inspect and modify, along with published evaluations and limitations so outsiders can challenge safety claims. He supported independent testing and accountable enforcement and said models should be withheld only when evidence indicates their release would materially increase catastrophic risks that narrower measures cannot address.

Analysts and other organizations offer different emphases. A recent Atlantic Council analysis suggested that competitive pressures make enforceable safety standards necessary, while noting that U.S.–China mistrust complicates multinational agreements. Observers have also flagged antitrust issues around cooperation: OpenAI has asked lawmakers whether rival developers could legally agree to slow development, highlighting tensions between coordination for safety and competition law constraints.