A group of crypto companies, including Coinbase, Block, and BitGo, has signed a letter to leading AI labs arguing that safety guardrails on frontier models are blocking legitimate security work. The letter, backed by dozens of other firms, contends that attackers face no such limits from the same guardrails.
The Core Complaint
The letter's central argument is that the safety measures built into advanced AI systems are hampering security researchers and other legitimate users. These guardrails, designed to prevent misuse, are also blocking efforts to test, probe, and defend against real threats. Meanwhile, the letter says, adversaries face no such limits — they can exploit the same models without worrying about safety constraints.
Who Signed On
Coinbase, Block, and BitGo are among the signatories, joined by dozens of other companies from the crypto and digital asset space. The letter does not name the specific AI labs it's addressed to, but it's aimed at the developers of frontier models — the most powerful and capable AI systems.
The Asymmetry Problem
The letter highlights an asymmetry: defenders are forced to work within safety guardrails, while attackers operate without them. That means security professionals can't fully test models for vulnerabilities, while malicious actors can freely probe for weaknesses. The signatories argue this puts the entire ecosystem at risk.
The letter is a direct appeal to AI labs to reconsider how they implement safety measures. The companies behind it are not asking for guardrails to be scrapped, but they're making the case that a blanket approach doesn't work for security work. Without the ability to stress-test models, they argue, the very systems meant to be safe become more vulnerable.
The letter is now in the hands of the AI labs. Whether they'll adjust their guardrails in response is the next question.
