“We've fixed that” is a claim, not a fact.
When a company tells us it has changed or clarified its position, that reply is printed here word for word, and then checked against the archive. A claim counts as confirmed only when the monitor holds a real before-and-after of the document the company named, captured on or after the day we wrote to them. Anything short of that is listed as unverified, with the reason.
“Regarding whether Krea uses customer data to train AI models, Krea states it does not. For Business and Enterprise plans, Krea commits that Inputs and Outputs are not used to train its AI models. [...] If you were looking specifically at the general Terms of Use and Privacy Policy, those pages do not spell this out. The training policy is published in the Business Terms and the Enterprise page above.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“It is not accurate to state that “Chegg trains its own and third-party data AI models on what students submit.” While the policy language that you quoted is accurate, it is not complete. Chegg’s policy has always been that it does not use content submitted by users Chegg knows to be under 18 to train or improve any artificial intelligence model, and Chegg does not disclose content from such users to third parties for their own purposes. [...] we have since made that clear, even though this protection was always a part of Chegg’s practices.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“For services for individuals, such as ChatGPT and Codex, OpenAI may use content to train our models. Users can opt out of model training by either: Turning off “Improve the model for everyone” under Settings → Data Controls in ChatGPT; or Selecting “Do not train on my content” through our privacy portal. [...] After opting out, new conversations will not be used to train our models. [...] For our business offerings, including ChatGPT Business, ChatGPT Enterprise, and the API Platform, OpenAI does not train on inputs or outputs by default.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“Thank you for your question. MyFitnessPal uses tools like artificial intelligence and machine learning systems to enhance and personalize our Service. However, we have not and are not using your personal information to train any proprietary models at this time. For more information, please see the How We Use Personal Information section in our Privacy Policy.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“I have actually now added this line into our privacy policy. You can see this in the section "3rd Party AI connections".”document they named ↗
4 paragraph(s) changed on the document they named, captured before and after.
see the before/after diff →“Enterprise Accounts: Customer data on Enterprise plans is strictly excluded from AI model training. Non-Enterprise Accounts: Data on non-Enterprise plans may be used to improve general AI models, but users can opt out at any time by contacting privacy@heygen.com. This opt-out applies retroactively to previously processed data. Third-Party Subprocessors & Vendors: All third-party AI vendors and subprocessors integrated with HeyGen operate under Zero Data Retention (ZDR) policies and are prohibited from training models on customer data.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“Thank you for reaching out and reviewing Bitwarden! Correct, Bitwarden cannot access Vault data.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“To clarify, D-ID does not use your resources or personal information to train, fine-tune, or improve our AI/ML models, or for any other purpose unrelated to providing our services. As mentioned previously, the resources you upload to your D-ID account, including images, videos, audio, and other user-provided content, are used to provide the services you request and are not accessible to others.”document they named ↗
The document was under watch across this period with a before-state on file, and the monitor recorded no change to it.
“Synthesia does not develop large language models or general purpose AI models. [...] We pre-train these AI components on performances by paid actors we commission, on performances that are publicly available, and on performances we license. We do not pre-train them on customer data. [...] Beyond that consented fine-tuning step, Synthesia does not use any other customer data, including the inputs to the product or the videos produced with it, to pre-train its AI models.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“Hugging Face does not store any user data for training purposes. Please refer to our Privacy Policy: https://huggingface.co/privacy or Security & Compliance: https://huggingface.co/docs/inference-providers/en/security for further information.”document they named ↗
This monitor holds only one capture of the named document, so there is no second version to compare it against. The gap is in our coverage, not necessarily in their account.
“Recraft uses user-generated content—such as images created or uploaded, and text entered in prompts or chat mode—to improve its AI models by default. However, there is an opt-out control available: On the Pro plan, users can opt out of model training by turning off the "Help improve Recraft" toggle in their Profile settings. [...] Training opt-out is not available on the Free plan. For Team plans, opt-out is applied by default. API inputs and outputs are never used for model training.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“We do not have that opt-out information published on a page at the moment [...] Currently, opt-outs are supported for paid accounts only. Opt-outs are respected for customers who sign up for Free Trials or Proof of Concepts (PoCs) of specific features or functionality through their paid accounts. [Asked whether a standalone free trial, not taken out under a paid account, has any route to opt out:] You are correct, the answer is no.”
The company described its position in correspondence but pointed to no public document. There is nothing to diff, and nothing a reader can check independently of their word.
“Descript does not use any user data for AI model training unless the user has explicitly opted in via the "Share data with Descript" setting. [...] For Enterprise drives, data sharing is disabled by default and cannot be enabled. No user data is shared with sub-processors (such as OpenAI) for the purpose of training their AI models.”document they named ↗
The document was under watch across this period with a before-state on file, and the monitor recorded no change to it.
“So yes: my earlier phrasing of "AI Data Usage toggle" was not the best label for the published control. The current help-center wording uses "AI Data Retention" [...] On the retention point, your reading matches the published guidance. I do not have a separate route to name for removing training data that was already collected before the opt-out. The available control applies going forward.”document they named ↗
1 paragraph(s) changed on the document they named, captured before and after.
see the before/after diff →“I lead Legal at Glean Technologies Inc. Glean Technologies Inc. does not train on customer/user data, this is clearly stated in our security documents, which are confidential.”document they named ↗
The document was under watch across this period with a before-state on file, and the monitor recorded no change to it.
What would move one of these to confirmed
The monitor re-checks every tracked document each week. Once it holds a second capture of a document named above, any later rewrite produces a real before-and-after, and the entry moves up on its own. Nobody marks it by hand. Entries whose reply named no document at all can't be confirmed this way, and are expected to stay where they are.
the full right-of-reply register →