STORY · FORSKNING_
Researchers demonstrate self-replicating AI virus; compute becomes costlier as AI grows smarter
Researchers from University of Toronto, Vector Institute, University of Cambridge and ServiceNow have built a proof-of-concept for a data virus that uses open-weight LLMs to identify vulnerabilities, exploit them and replicate itself on compromised systems. The virus achieved an 80% success rate on vulnerability detection and 88% on self-replication, with a combined 37% full-attack success rate.
WHY IT MATTERS
The demonstration shows that autonomous, self-sustaining AI threats are no longer theoretical, and suggests that future internet infrastructure will become an 'ecology' of attacking and defending AI agents. At the same time, AI compute is noted to become significantly more expensive as models approach human-equivalent capabilities.
SOURCES
MACHINE-GENERATED SUMMARY This summary is written by machine from the sources below. We sort and explain — but we are a way into the field, not the final word. Check the source when something matters to you.