Vitalik: Supports restricting the training of advanced cyberattack capabilities for Kimi 3.1

on October 26, Vitalik Buterin stated he supports proactively excluding top offensive cybersecurity capabilities from Kimi 3.1's training data, arguing that "limiting use cases" is far preferable to "limiting users."

He noted that society still needs time to prepare for more powerful AI cyberattack capabilities, including strengthening system security, auditing legacy code, and conducting rigorous formal verification on new code. Vitalik warned that if the first wave of major AI risks are mishandled, the future could trend toward highly closed models, comprehensive KYC thresholds, and mass surveillance, with different users potentially receiving varying levels of model access based on their identity, credentials, or nationality.