Authored by David Rolnick, Jonathan Hayase, Eric Wallace, Nicholas Carlini, Arthur Conmy, Thomas Steinke, Matthew Jagielski, Florian Tramer, Krishnamurthy Dvijotham, Daniel Paleka, Katherine Lee, Milad Nasr, A. Feder Cooper

In this whitepaper, the authors introduce the first model-stealing attack that extracts precise, nontrivial information from black-box production language models like OpenAI’s ChatGPT or Google’s PaLM-2. Specifically, their attack recovers the embedding projection layer (up to symmetries) of a transformer model, given typical API access. For under $20 USD, their attack extracts the entire projection matrix of OpenAI’s ada and babbage language models. They thereby confirm, for the first time, that these black-box models have a hidden dimension of 1024 and 2048, respectively. They also recover the exact hidden dimension size of the gpt-3.5-turbo model, and estimate it would cost under $2,000 in queries to recover the entire projection matrix. They conclude with potential defenses and mitigations, and discuss the implications of possible future work that could extend this attack.