English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The generative Artificial Intelligence (AI) tools based on Large LanguageModels (LLMs) use billions of parameters to extensively analyse large datasetsand extract critical private information such as, context, specific details,identifying information etc. This have raised serious threats to user privacyand reluctance to use such tools. This article proposes the conceptual modelcalled PrivChatGPT, a privacy-preserving model for LLMs that consists of twomain components i.e., preserving user privacy during the datacuration/pre-processing together with preserving private context and theprivate training process for large-scale data. To demonstrate itsapplicability, we show how a private mechanism could be integrated into theexisting model for training LLMs to protect user privacy; specifically, weemployed differential privacy and private training using Reinforcement Learning(RL). We measure the privacy loss and evaluate the measure of uncertainty orrandomness once differential privacy is applied. It further recursivelyevaluates the level of privacy guarantees and the measure of uncertainty ofpublic database and resources, during each update when new information is addedfor training purposes. To critically evaluate the use of differential privacyfor private LLMs, we hypothetically compared other mechanisms e..g, Blockchain,private information retrieval, randomisation, for various performance measuressuch as the model performance and accuracy, computational complexity, privacyvs. utility etc. We conclude that differential privacy, randomisation, andobfuscation can impact utility and performance of trained models, conversely,the use of ToR, Blockchain, and PIR may introduce additional computationalcomplexity and high training latency. We believe that the proposed model couldbe used as a benchmark for proposing privacy preserving LLMs for generative AItools.

Privacy Preserving Large Language Models: ChatGPT Case Study Based Vision and Framework