Is Claude safe to use?
AI assistant from Anthropic. Everyday conversation, writing, analysis, code, and long documents, plus Artifacts: side-panel documents and small interactive tools you can edit, download, and optionally publish.
Last reviewed: 5 August 2026
Does Claude train on what you type?
Yes, unless you switch it off — and you can.
Can you turn Claude's training off?
Model improvement: Settings → Privacy → turn off Help Improve our AI models. If it is already off, leave it off. Use Incognito chats for anything you would not want in a training pipeline; Anthropic says Incognito chats are not used for model improvement even if Model Improvement is enabled. Avoid thumbs-up/down feedback on sensitive conversations, because feedback can store the related conversation for up to five years. Public share links: if you have ever clicked Share, go to Settings → Privacy → Manage next to Shared chats, then unshare anything you would not put on a bulletin board. You can also open the chat, use the Share menu, and change visibility from Public to Private. For a published Artifact, open it and choose Unpublish. Anthropic says unpublishing an Artifact is one-way for that artifact version, so do it deliberately. The durable rule is simple: if you would not post it on a public board, do not create a public share link. Paste a private copy, export a file, or sit next to the person instead.
Is what you put into Claude private?
Anthropic collects what you type and upload into Claude, plus Claude's replies, on consumer Free, Pro, and Max accounts, including Claude Code used with those accounts. Anthropic's current Privacy Center says model-improvement use of those chats happens if you allow Model Improvement, if a conversation is flagged for safety review, or if you otherwise explicitly opt in. Its legal privacy policy still says inputs and outputs may be used to train and improve models unless you opt out. The practical advice is the same either way: check Settings → Privacy, keep Model Improvement off if you do not want chats used for training, use Incognito chats for sensitive work, and avoid thumbs-up/down feedback on sensitive threads. Feedback can keep the related conversation for up to five years. Claude for Work and API usage follow separate commercial terms. Separate from training: if you use Share on a chat, or Publish on an Artifact, on a Free/Pro/Max account, Claude creates a public link that anyone who has it can open. A shared chat is a snapshot of the conversation and any Artifacts up to the moment you shared it; later messages stay private unless you share again. Attached files and raw MCP tool-call data are not included in shared chat snapshots, but the visible conversation and Claude's final outputs are. It is not a locked family Google Doc. In July 2026, TechCrunch reported that shared Claude chats and Artifacts had appeared in search results after public links were posted where search engines could crawl them, with other outlets reporting examples that included medical details, internal work documents, and children's contact information. Anthropic says it does not provide chat directories or sitemaps to search engines and that shared links are not guessable; the risk is the public link itself. Team and Enterprise sharing stays inside the organization. Review and revoke public shares in Settings → Privacy → Shared chats, and unpublish public Artifacts from the artifact itself.
So, is Claude safe to use?
Mostly, with care. We currently rate Claude "Use With Caution" (privacy risk: Medium). Worth reading the points above before you share anything personal.
The Skeptic's take
“I treated 'anyone with the link' like a private Google Doc for family. It is not. If I would not thumbtack it in the grocery store vestibule, I do not hit Share — and I just checked Settings → Privacy → Shared chats to be sure.”
— The Skeptic
The official documents behind this review
We read these so you don’t have to. Check them yourself any time.
Wondering how we decide these ratings? Read how we review AI privacy.
Keep reading
Knowing what’s safe is half of it.
The other half is knowing what to do with these tools. We teach that part — tutorials, prompts, plain English, no judgment. For less than a cup of coffee a month.
Show Me How This Works →Start free. Cancel anytime. No judgment, ever.