ChatGPT at work: how chatbots help employees, but threaten business
SecureList
OCTOBER 13, 2023
Given that LLMs are prone to so-called unintended memorization (memorizing unique sequences like phone numbers that do not improve the quality of the model, but create privacy risks) data that ends up in the training corpus can then be accidentally or intentionally extracted from the model by other users. Malicious client.
97 
Let's personalize your content