Crypto Giants Urge Removal of AI Guardrails to Boost Security Research
Cryptocurrency and blockchain firms including Coinbase, Block, and BitGo have signed an open letter warning that artificial intelligence safety guardrails on frontier models are hindering legitimate security research. According to CoinDesk, the coalition argues that defensive cybersecurity operations face strict operational limits while malicious actors encounter no such restrictions when developing attacks.
The letter points to a growing friction point between commercial AI development and digital asset security. Firms that rely on advanced language models and automated systems to scan for vulnerabilities claim that current restrictions make routine defensive testing difficult. At the same time, adversaries leverage unrestricted systems to scale malicious campaigns against digital infrastructure.
Security Guardrails Limit Defensive Operations
According to the signed letter detailed by CoinDesk, strict alignment and safety policies implemented by AI developers often misclassify standard penetration testing and vulnerability analysis as harmful activity. This classification leads to blocked queries and restricted model access for security professionals working to protect blockchain protocols.
Major industry participants across the financial technology and crypto sectors argue that these friction points weaken overall network security. While defensive teams must operate within rigid automated constraints, malicious actors utilize alternative or unaligned models that impose no such boundaries.
Industry Coalition Calls for Balanced Access
The signatories, which include prominent digital asset custody and exchange platforms, are pressing AI labs to reevaluate how safety filters interact with authorized security workflows. The coalition asks for specialized access or modified safety parameters that allow verified defenders to utilize frontier models for threat intelligence and code auditing.
As automated threats evolve across financial networks, the debate over AI access highlights a complex challenge for developers balancing the prevention of dual-use harms with the operational needs of enterprise cybersecurity teams.