… AI is data-hungry: platforms like ChatGPT and Gemini are built on wild amounts of data, both scraped freely from across the web and collected from people interacting directly with the AI. …
… Google is focusing on “client-side encryption” in which you hold the keys and can block access to sensitive data from anyone, even Google. Also, your company’s data isn’t used to train public AI models, and no one will see your spreadsheets unless you allow it. …
… While Sattelberg notes that Health Coach successfully referenced both the previous night’s sleep data and a workout from the day before, it just completely made up a 5-mile run that he never actually took. …
… Gemini chats and uploaded documents are used for training data, and opting out means you lose out on essential features like chat history. …
… This will delete everything, but keep in mind that OpenAI says it can take a month or so to truly remove the data from the servers and that some of it will remain in the system for security or legal reasons or if the data has already been used for training after de-identification. …
… While the research relied on explicit consent and encrypted data collection, any future consumer implementation would need strong safeguards. Google suggests that on-device processing and face authentication could help address those concerns. …
… If you’re producing work where IP ownership is a real concern, Firefly’s training data argument still holds. …
… To get around this, devices without GPS usually rely on their companion smartphone for positional data. …
… This is essentially a data-mining operation. …
… It should be an easy fix, and by adding it to the API, Google remains in control of the location data and battery use through Find Hub. …