Claude Updates Terms of Service: How Will Your Conversation Data Shape the Future of AI?

AI company Anthropic recently announced an update to its consumer terms and privacy policy for its AI assistant, Claude, giving users greater data control by allowing them to decide whether to allow their conversation content to be used for model training. This article will delve into the key points of this update, the specific impact on users, and the considerations behind Anthropic.

Recently, if you are a user of the AI assistant Claude, you may have noticed a new notification popping up within the application. This is not an ordinary update reminder, but an important policy adjustment from its developer, Anthropic, concerning how your conversation data is used and the control you have over it.

In short, Anthropic is giving users a clear choice: whether you are willing to let your conversations and code content be used to help train and improve future Claude models. This change aims to enhance the AI’s capabilities and safety while returning data control to the user.

Your Choice: Should Your Data Be Used for AI Training?

The core of this update is an “opt-in” mechanism. In the past, the data usage terms of many tech services were hidden in lengthy texts, but Anthropic has chosen a more transparent approach this time.

What does this mean? When you use the Free, Pro, or Max versions of Claude, you can decide for yourself whether to allow Anthropic to use your conversation history to train the AI. This setting can be changed at any time in your “Privacy Settings,” giving users great flexibility.

Who is affected?

Applies to: Individual users on Claude Free, Pro, and Max plans.
Does not apply to: Users of commercial services are completely unaffected. This includes Claude for Work (Team and Enterprise plans), Claude Gov, Claude for Education, and developers and businesses using the API through third-party platforms like Amazon Bedrock or Google Cloud Vertex AI.

This is a very important distinction. Anthropic explicitly states that the data of commercial customers will continue to be protected by its business terms and will not be used for model training, which is undoubtedly to protect the data privacy and business secrets of enterprise users.

Why Does Anthropic Need Your Data?

So, the question arises, why does Anthropic want you to share your data? Just as humans need to learn and interact to grow, large language models also need a large amount of real-world data for optimization. Every question you ask, every collaboration, can become nourishment to make Claude smarter.

Anthropic states that this data is primarily used for two key purposes:

Enhancing Model Safety: By analyzing real conversations, the system can more accurately identify harmful content, such as scams or abusive behavior. At the same time, it can also reduce false positives, avoiding incorrectly flagging harmless conversations as dangerous.
Strengthening Core AI Skills: Whether it’s helping developers debug code, performing complex data analysis, or providing more logical reasoning, real interaction data is an indispensable teaching material for training the AI to master these advanced skills.

From this perspective, this is not just one-sided data collection by Anthropic, but more like a collaborative relationship—users help build a more powerful and safer AI tool for everyone by sharing their data.

Data Retention Period Extension: A Shift from 30 Days to 5 Years

Accompanying the choice of data use is an adjustment to the data retention policy. This adjustment is also closely related to your choice.

If you choose to share your data: The retention period for your conversations and code records will be extended to five years.
If you choose not to: The data retention policy will remain at the current 30-day period.

Why is a five-year retention period necessary? Anthropic explains that the development cycle of AI models is very long; a model released today may have started development 18 to 24 months ago. Using a consistent dataset for training over such a long period helps to make the model’s responses, reasoning, and output more stable, ensuring a smoother experience during model upgrades.

Furthermore, a longer data retention period also helps to improve the classifier systems used to detect abusive behavior, allowing them to learn from longer-term data and more effectively counter spam or inappropriate use.

How Should Users Respond?

Faced with this update, whether you are a new or existing user, Anthropic provides clear guidance.

For existing users: You will see an in-app notification when using Claude asking for your preference. You can make a choice immediately or choose “Decide later” and complete the setup by September 28, 2025. After this date, you will need to make a choice to continue using Claude.
For new users: This choice will be part of the registration process, allowing you to set your preferences before you start using the service.

The most important point is that this is not a permanent decision. You can change your option at any time by going to the “Privacy Settings” page.

What If I Change My Mind?

This is a question many people are concerned about. If you initially agree to share your data but later change your mind, what happens?

You can go to “Privacy Settings” at any time to turn off the model training option. Once turned off, Anthropic will no longer use any of your new conversations or code with Claude for model training.

However, it is important to note that data you have previously allowed to be used for training may still exist in models that have already started or completed training. But the system will stop using your previously stored conversations for any new training in the future. Similarly, if you delete a specific conversation, that conversation record will not be used for future model training.

In summary, Anthropic’s latest policy update attempts to find a balance between advancing AI technology and protecting user data privacy. It gives users control over the decision to use their data and communicates the motives behind it more transparently, which may also provide a valuable reference for the future development of the AI industry.