Using AI

willowbark · ‎07-20-2025

Privacy Policy Comparison Table

Platform Data Usage Opt-out Options Data Retention Transparency Special Notes

OpenAI (ChatGPT)	User data is generally used for training unless opted out.	Users can opt out, but it's not the default.	Retains data indefinitely unless manually deleted; temporary mode available (30 days).	Clear, readable policies on data practices.	Users must actively manage preferences to limit usage.
Anthropic (Claude)	Conversations may be used for training by default; option to opt in or out.	Opt-out available but has changed to prompt users for consent.	Retained for 30 days if opting out; training data can be kept longer (~5 years).	More straightforward but still has privacy concerns.	Previously stronger privacy stance, now more flexible.
Google (Gemini)	Uses data for model training and performance improvement; extensive data sharing.	Complex opt-out process primarily compliant with GDPR.	Retention policies are often vague and dependent on region.	Policies are broad and often unclear regarding specific user interactions.	Metadata and location data often collected; data used in ad targeting.
Microsoft (Copilot)	Similar to Google; collects broad user interactions for improvements and data analytics.	Opt-out options exist but are also convoluted.	Retention practices similar to Google; often lacks clarity.	Privacy policies are integrated across services, making specifics harder to find.	Often shares data with other Microsoft services and third parties.
Meta (Meta AI)	Aggressive data collection from public and user-generated content; broader categories included.	Limited and often unclear opt-out procedures.	Retention policies can be lengthy; relies heavily on user data.	Transparency is lacking; complex language in policies.	Users may not be fully aware of how their data is utilized, esp. in training models.

Analysis

Data Usage and Training: Most platforms utilize user-generated data to improve their models. OpenAI and Anthropic provide users with options to limit or opt-out of certain data uses, but this is often not enabled by default.

Opt-Out Options: Platforms like Anthropic have shifted their policies to require active user consent for data training, reflecting changing attitudes toward privacy. In contrast, platforms like Meta offer limited and unclear options, making it hard for users to navigate privacy controls.

Data Retention Practices: Retention policies vary widely, with OpenAI retaining information indefinitely unless expressly deleted. Google and Microsoft also retain data for analytics but do not provide clear pathways for users to understand or manage retention.

Transparency: OpenAI is noted for its clear communication regarding data practices compared to others like Meta and Google, where vague policies often leave users confused about data handling.

Special Notes: Meta stands out for its aggressive data collection techniques. On the other hand, Apple's AI model prioritizes privacy by processing data minimally and locally, thus not being included in this comparison but noted for its strong stance on user privacy.

source: Duck.ai

willowbark · ‎07-20-2025

Something I have discovered by using Duck.ai (which is the only AI I have ever used) and that is discovering inaccuracies in its answers and giving me extraneous information that I never mentioned to it or asked it.

For example

After telling Duck.ai where I reside I mentioned a specific law that had been changed and was in the news and online too of course. Yet it answered my question based on old information before the law was changed, When I told AI that no, it was incorrect it

apologized and then produced the information I was seeking based on the new law and there had been copious amounts of that online. Very disappointing.

Example #2

Based on where I reside which I reiterated to Duck.ai, I asked about my property taxes and what percent can be written off on my Federal Income Tax. The response I got specified the filing status for married people but I never asked it about filing status and it failed to include single filing status leaving me to believe that its programmers were biased toward married people only owning homes and not a single person.

When I brought that to its attention, it again apologized and included the information for single fliers as well.

I then asked it why it was relevant to even bring that up at all since filing status was never part of my question. My question was merely asking what percent can I write off from my property taxes.

I received another apology. The aswer was pretty straightforward and did not have to include filing status.

By the way, I already knew the answer just wanted to see what it would say.

tansy · ‎01-02-2011

I am attending a lecture on the subject this coming week.

Laura14 · ‎06-09-2014

ChatGPT is very clear. If you are logged in to save your chats (you don't have to be to use it), you can delete and it tells you straight up it will be gone from the server in 30 days or longer depending on what it is.

willowbark · ‎07-20-2025

@Laura14 wrote:
ChatGPT is very clear. If you are logged in to save your chats (you don't have to be to use it), you can delete and it tells you straight up it will be gone from the server in 30 days or longer depending on what it is.

I don't use AI on my phone and while ChatGPT is available from what I understand on Microsoft, I don't use their browser.