Patrick Zandl

Interesting thread, but several claims need correcting: • You cite the same Stanford study 5x from different sources to pad your references. That’s not thorough research, that’s decoration. • “No clear opt-out” for Google Gemini? in Settings/Activity takes 10 seconds. • You lump consumer free tiers with API/Enterprise — OpenAI API and Anthropic API don’t use data for training at all. For someone advising “Fortune 500 clients,” this distinction matters. • The claim Google could “deduce health conditions from recipes” and share with insurers isn’t in the Stanford study. That’s your speculation dressed as their finding. • Anthropic’s policies differ significantly between claude free, Pro, and API. Saying they “flipped to opt-out” without this context is misleading. • Zero concrete cases cited where LLM training data leaked back to identify a specific user. The risk is real, the evidence you present is not. The core issue (default data collection, weak US regulation) is valid and important. But mixing facts with speculation and self-promotion undermines the message you’re trying to deliver.