They are still training on your data (trustedrouter.com)

A recent Google DeepMind paper reveals that AI companies are still leveraging user data, despite claims of data privacy and zero retention. The research introduces a technique called Generative Data Refinement, which transforms private or unusable data into valuable training data by rewriting sensitive portions while retaining useful information. This method addresses the ongoing data shortage in training datasets, which often lack diverse and current content found in personal communications and proprietary documents. The implications for the AI/ML community are significant, as this approach allows companies to create "grounded synthetic data," ensuring high recall and precision while sidestepping legal and ethical concerns related to using real data. The findings stress the importance of scrutinizing data usage policies, as companies can still train models on synthetic versions of user inputs, raising questions about transparency and consent. Solutions like TrustedRouter and its privacy-focused frameworks emphasize the need for verifiable data handling, highlighting a pressing demand for clarity in the evolving landscape of AI training practices.