In plain English
This page explains how AI systems can change over time through updates, tests, retraining, memory, and approvals even when no single model rewrites itself.
- Why this matters: AI risk can come from the whole arrangement, not one obvious model.
- What to look for: data, memory, routes, adapters, tools, evaluators, updates, and rollback paths.
- Technical version below: the expert terminology remains available and is linked through the glossary.
Open-Ended Evolution Without Autonomy
The report corpus expands the earlier CognivirusA behavior pattern that can survive, move, or reappear across a changing AI system. Open glossary definition thesis: variation, selection, inheritance, and succession can be properties of the development process even when no deployed model self-modifies at runtime.
Pipeline-level evolution
A CI system can generate candidate prompts. A fine-tuning job can create adapters. A merge tool can recombine checkpoints. A canary process can select winners. A registry can preserve descendants. Humans may approve every step, but the pipeline still creates selection pressure.
Why this matters
Treating “not autonomous” as “not evolutionary” misses the risk. The ecology may still preserve behaviors through repeated derivative creation and selection.
Boundary
This page does not collapse ordinary engineering automation, governed model breeding, and uncontrolled replication into one category. It distinguishes them so controls can match the actual risk.