Coinbase CEO Brian Armstrong has rejected calls for a new AI self-regulatory body, arguing that existing laws already provide sufficient protection against harmful AI. The proposal from Google DeepMind CEO Demis Hassabis suggested a federally overseen standards body for testing frontier AI models. Armstrong countered that such bodies create dual approval processes and that regulation should not be designed around hypothetical problems. Meanwhile, Coinbase’s head of platform Rob Witoff revealed that the crypto exchange now writes over 95% of its code using AI, more than doubling the 40% figure reported in February. The company cut 14% of its workforce in May to reorganize around AI, though adoption has caused setbacks including an erroneous match notification.
Coinbase CEO Brian Armstrong has rejected calls for a new AI self-regulatory body, arguing that existing laws already provide enough protection against harmful AI. His remarks followed a July 14 proposal from Google DeepMind CEO Demis Hassabis, who called for a federally overseen standards body to test and certify frontier AI models before deployment. Hassabis stated that artificial general intelligence could arrive within a few years and may introduce cybersecurity, biological, and national security risks.
Tech entrepreneur Chamath Palihapitiya described the framework as “quite well reasoned,” while OpenAI CEO Sam Altman called it “a thoughtful proposal.” Microsoft CEO Satya Nadella labeled it “an important piece,” adding that the goal should be to avoid “any model that breaks the world.” However, Armstrong disagreed, maintaining that such arrangements create a dual approval process forcing businesses to satisfy both state regulators and industry bodies.
“Why design regulation around a hypothetical problem,” Armstrong wrote, adding that existing laws preventing fraud and awarding damages already provide broad protections. He also noted that AI developers have a strong commercial incentive to release safe products, since users avoid dangerous tools.
The debate comes as Coinbase deeply integrates AI into its operations. Rob Witoff, the company’s head of platform, told Cointelegraph that between 95% and 100% of the exchange’s code is now written by or with large language models, more than doubling the roughly 40% estimate from February. In May, Coinbase cut 14% of its workforce to reorganize around smaller, AI-centered teams.
AI usage varies by task: sensitive areas like cryptography still receive detailed human review, while internal prototypes are built almost entirely through automation. Other crypto firms including Gemini, Crypto.com, Kraken, Messari, and Dune have also reduced headcounts while expanding AI use. However, rapid adoption has led to setbacks, such as an AI-generated notification that incorrectly reported a FIFA World Cup match result before the match started.
