All llm api providers should be compelled to return a checksum-like proof of quantization level of the model that served the request. Basic transparency should be the bare minimum.