Ben has been loving Anthropic’s AI tool, Claude, for months, but with its recent privacy changes, their relationship is beginning to strain. Claude, and plenty of other platforms utilising AI models including LinkedIn, have begun sneakily changing their privacy terms in order to train their AI models using your data. At Asenda Law, we’re here to inform you about what the law says, and what steps you should take to stay in control.
What’s changed, and what do these AI privacy changes mean for your data?
Anthropic recently updated its Consumer Terms and Privacy Policy so that Claude can use your chats and coding sessions to train future models, if you allow it. Existing users of the platform prior to the updated terms were asked to make a choice by 28th September 2025, and new users who have signed up since then are now having to choose at sign-up. Anthropic is also extending data retention to five years if you permit training. If you don’t accept, their usual 30-day retention continues. Only new or resumed chats are in scope, and deleted conversations aren’t used for future training.
Who is and isn’t affected by Claude’s new AI privacy terms?
This does not apply to services under Anthropic’s Commercial Terms, such as Claude for Work/Team/Enterprise, Government, and Education, or the API, including via Bedrock or Vertex. Those remain under their existing contractual privacy protections. However, we’ve seen that a lot of users are using Claude under the consumer terms for work purposes too, so business owners should be very careful with what sensitive data they choose to input into AI platforms.
Can you change your privacy settings?
Luckily, the answer is yes. In Claude, you’ll see a window with a prominent “Accept” button and a friendly “Help improve Claude” toggle which allows you to change your data privacy settings. Similarly, LinkedIn has default toggled the permission to use your posts and data to train on, and so if you don’t agree you need to manually update this preference setting.
UX folks will recognise the online choice architecture at play. Under UK GDPR regulations, valid consent must be freely given, specific, informed, and unambiguous. UK regulators, including the ICO and CMA, have cautioned against designs that steer users towards more data sharing, which Claude and other platforms stealthily switching to this manual opt-out approach are arguably toeing the line on.
What steps should individuals take to protect themselves against AI privacy changes?
Decide whether you want to allow your data to be used in training AI models, and if not, update your data privacy settings. You can change your choice if you change your mind later on, but it’s important to note that data that’s already used in training AI models can’t be removed, so make your choice accordingly.
Don’t resume old chats you don’t want in scope for AI training.
For confidential matters or users, switch to a commercial plan or follow the API route. Better yet, avoid inputting sensitive data altogether.
This is a clear shift from “privacy-by-default” for consumers to a consent-based model that supports model development. That’s not inherently unlawful, but it does put the onus on users to make, and evidence, an informed choice – and it raises familiar UK GDPR questions about consent quality and interface design.
It’s disappointing to see Anthropic, LinkedIn, and a host of other platforms shifting their stance on this, as providers in this space can really set themselves apart by taking a transparent stance on how they use data in their AI training.
What should you do moving forward?
The safest operating assumption remains: treat prompts as potentially persistent, and pick the settings or contracts that match your risk profile.
If you’re concerned about what these changes might mean for you or your data, get in touch with us at Asenda Law. We’ve got all the knowledge and expertise to help you navigate these changes in a way that keeps your IP secure, and puts your mind at rest.


