Migliaccio & Rathod LLP is investigating whether users of Claude clearly understood and meaningfully consented to the use of their chats, coding sessions, and other interactions for training Anthropic’s AI models after Anthropic changed its consumer data practices in 2025.
Historically, Anthropic did not generally use ordinary consumer Claude conversations to train its models unless users provided feedback or otherwise expressly opted in.
In 2025, Anthropic changed that approach. For Claude Free, Pro, and Max users—and consumers using Claude Code through those accounts—Anthropic introduced a setting allowing users to permit their chats and coding sessions to be used to improve Claude. Anthropic also extended retention of participating users’ data to five years.
What Information May Be Used?
Anthropic states that, when a consumer chooses to allow model improvement, the data potentially used can include:
- Claude conversations;
- Claude Code sessions;
- content contained within those conversations;
- custom styles;
- conversation preferences; and
- certain information generated while using Claude for Chrome.
Anthropic states that raw content from connected services such as Google Drive is not automatically included simply because it is available through a connector, although information copied directly into a Claude conversation may be included.
What We Are Investigating
Migliaccio & Rathod LLP is investigating whether the choice presented to existing Claude users was sufficiently clear and whether users understood all of the consequences of allowing model training.
We are investigating whether:
- existing Claude users clearly understood that their chats could begin contributing to model training;
- the training choice was presented in a neutral and understandable manner;
- users understood that allowing training also meant extending relevant data retention to five years;
- users understood which new or resumed conversations could be used;
- the distinction between ordinary use, safety review, feedback, and model training was sufficiently clear;
- users who later changed their preference understood what happened to data already used in a training run; and
- consent was truly informed where users had to make a selection before continuing to use Claude after the transition period.
Anthropic states that existing users were shown an in-app notification and were required to select a training preference by October 8, 2025 to continue using Claude. Anthropic also states that the policy applies to new or resumed chats, rather than dormant historical chats that users never reopened.
What Happens If a User Changes Their Mind?
Anthropic currently states that users can change their model-training preference at any time.
If a user turns the setting off, Anthropic says it will stop using new chats and coding sessions for future model training and will stop using previously stored chats in future training runs. However, information already included in training that has begun, or in models that have already been trained, cannot simply be removed from those completed training processes.
Anthropic also states that deleting an individual conversation prevents that conversation from being included in future training.
Potential Claims May Include:
- unfair or deceptive trade practices;
- inadequate disclosure concerning model training and retention;
- privacy-related claims;
- breach of contract or related contractual claims; and
- unjust enrichment where consumer interactions were used to improve commercially valuable AI models.
Signs You May Be Affected:
- You used Claude Free, Pro, or Max before September 2025;
- you regularly used Claude for personal conversations, writing, analysis, or coding;
- you were presented with a model-training choice in 2025;
- you did not understand that allowing training would extend relevant retention to five years;
- you felt you had to make a choice in order to continue accessing Claude;
- you initially allowed model training and later changed your preference;
- you used Claude Code through a personal Claude account;
- you discussed personal, confidential, creative, or commercially valuable information with Claude; or
- you have screenshots or emails showing how Anthropic presented the new training choice.
This investigation concerns consumer Claude accounts. Anthropic states that the training policy does not apply in the same way to commercial services such as Claude for Work, the Anthropic API, Amazon Bedrock, or Claude through Google Cloud Vertex AI.
If you used Claude and believe your chats or coding sessions may have been used to train Anthropic’s AI models without your informed consent, we would like to hear from you.
Please complete the contact form on this page, send us an email at [email protected], or give us a call.