Coinbase CEO Brian Armstrong opposes new AI regulatory bodies, arguing existing laws are sufficient and market incentives drive safety.
Regulation & Gov ·
Coinbase CEO Brian Armstrong has argued against establishing a new AI self-regulatory organization, contending that current legal frameworks already offer sufficient safeguards against AI-related harms. Armstrong's position emerged after a proposal from Google DeepMind's Demis Hassabis for a federally overseen standards body to test and certify advanced AI models before deployment. Armstrong countered that such dual-layer approval systems—combining state regulation with industry bodies—would be redundant, noting that tort law, fraud prevention statutes, and unfair-practice rules provide broad remedies if frontier AI labs release harmful systems. He further argued that commercial pressure alone incentivizes safety, since users naturally abandon tools they perceive as dangerous.
The crypto exchange disclosed that between 95% and 100% of its code is now generated by or created with large language models, more than doubling a prior estimate of roughly 40% disclosed in February. This expansion follows a May workforce reduction of 14%, during which the company reorganized to center smaller, experienced teams around AI-driven operations. Coinbase's head of platform indicated that sensitive functions like cryptography still receive detailed human review, though internal prototypes can be built largely through automation.
What remains unclear is whether Armstrong's confidence in market-based safety holds across the broader crypto industry as firms accelerate AI adoption. Coinbase itself faced a setback when an AI-generated notification incorrectly reported a FIFA World Cup result, highlighting the tension between rapid deployment and accuracy.