Mirai

Anthropic's Frontier Red Team published research suggesting small AI models may be only months away from replicating across insecure devices as autonomous botnets. Experts told Engadget the capability is plausible, but that the framing overstates how new the threat is.