Anthropic warns ‘AI worms’ months away from self-replicating device attacks

The notable shift here is not the worm itself but who is flagging it: a frontier AI lab pre-announcing a capability threshold, which reframes autonomous replication as an imminent operational risk rather than a research curiosity. The exposure runs well beyond crypto and RWA desks, since the same always-on, lightly secured consumer hardware that anchors tokenized real-world assets is exactly the surface described. Whether this stays a lab warning or becomes a compliance and insurance line item for device-dependent tokenization projects is the open question worth tracking.