Compute Scaling

Increasing performance via more data.

Scaling law optimizing compute vs data.

Cost to run models in production.

Cost of model training.

Extracting system prompts or hidden instructions.

Compromising AI systems via libraries, models, or datasets.

Embedding signals to prove model ownership.

Detecting unauthorized model outputs or data leaks.