Counterfit

Counterfit is Microsoft's earlier command line tool for assessing the security of machine learning models: adversarial examples, evasion and model inversion attacks against classifiers and vision models, wrapping libraries like ART and TextAttack. It predates the LLM wave and targets traditional ML.

It's in maintenance mode; PyRIT is where Microsoft's effort went. Useful if your risk is adversarial attacks on a classifier.