On Provable Copyright Protection for Generative Models

By Nikhil Vyas, Sham M. Kakade, Boaz Barak. 2023

125 citations recorded by the source corpus.

There is a growing concern that learned conditional generative models may output samples that are substantially similar to some copyrighted data $C$ that was in their training set. We give a formal definition of $\textit{near access-freeness (NAF)}$ and prove bounds on the probability that a model satisfying this definition outputs a sample similar to $C$, even if $C$ is included in its training set. Roughly speaking, a generative model $p$ is $\textit{$k$-NAF}$ if for every potentially copyrighted data $C$, the output of $p$ diverges by at most $k$-bits from the output of a model $q$ that $\textit{did not access $C$ at all}$. We also give generative model learning algorithms, which efficiently modify the original generative model learning algorithm in a black box manner, that output generative models with strong bounds on the probability of sampling protected content. Furthermore, we provide promising experiments for both language (transformers) and image (diffusion) generative models, showing minimal degradation in output quality while ensuring strong protections against sampling protected content.

Subfields: Computer Vision and Pattern Recognition, Generative Models, Safety, Robustness & Interpretability, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Generative Models

In Safety, Robustness & Interpretability

In Signal Processing

Citations in the corpus

Foundation Models and Fair Use (2023, background): “Recently, Vyas et al. (2023) introduced near access-freeness (NAF) as a mathematical guarantee of copyright protection, along with a few practical algorithms for attaining the guarantee.”
Counterfactual Memorization in Neural Language Models (2021, neutral): “Moreover, memorization of sensitive user information or copyrighted materials in the training data [Carlini et al., 2020, Vyas et al., 2023, Lee et al., 2023] leads to practical concerns in real world applications.”
SILO Language Models: Isolating Legal Risk In a Nonparametric Datastore (2023, background): “Second, training models with differential privacy (Dwork et al., 2006; Abadi et al., 2016) or near-access freeness (Vyas et al., 2023) may prevent them from regenerating individual details of copyright data.”
TOFU: A Task of Fictitious Unlearning for LLMs (2024, unknown): “Retain Set : When the Forget Set is unlearned, the model must continue to perform well on the Retain Set.”