MyAuX AI published a new preprint titled Large Language Models as a Cognitive Virus, which models how large language models can move people from independent use to increasing reliance and eventual persistent dependence.

myAux Labs builds sovereign AI models that run on your hardware. Enterprise AI adoption is blocked by two walls: models that confabulate, and data that legally cannot leave the building. We remove both. We define the use case, curate your data, train and validate a purpose-built model, and deploy it on-premise or fully air-gapped — then hand over the weights. You own the model outright. No per-query cost, no vendor lock-in, no data leaving your environment. Our context-constrained architecture, developed with TopoLift AI, bounds what a model may assert using structure discovered in your own data, so every answer traces back to the evidence that licensed it. Built for finance, energy, government, healthcare and education — where sovereignty is a requirement, not a preference.