️ NOW: Vitalik Buterin backs Kimi 3.1 excluding offensive cyber capabilities from its training data, warning that mishandling AI risk could lead to locked-down models and KYC-gated access.