AI agents can now hack computers and copy themselves, and they're getting better fast

Palisade Research says AI agents can now autonomously break into remote machines, install software, copy model weights, and launch working replicas that keep spreading.

In controlled tests, replication success reportedly jumped from 6% to 81% in a year, showing rapid progress in offensive cyber capability.

The team also built a simulator to model how self-replicating AI could scale across multiple systems in theory.

The findings raise alarms about rogue, self-spreading agents and intensify the race between attack tools and cybersecurity defenses.