Can you use Claude output commercially?

Commercial use is allowed on every Claude plan including Free. What the Consumer and Commercial Terms say about output ownership, and what is restricted.

Yes. You can use Claude's output commercially on every plan, including Free. Anthropic's Consumer Terms assign users “all of our right, title, and interest—if any—in Outputs”, and the Commercial Terms governing Team, Enterprise and API use say the customer “owns its Outputs”. There is no personal-use-only restriction on any tier. The one universal prohibition is using outputs to train a competing AI model.

What follows is the clause language itself, the restrictions that do apply, how data-training rules differ between consumer and commercial plans, and where the law is genuinely unsettled.

#What do Anthropic's terms actually say about ownership?

Two different documents govern two different sets of plans, and they use meaningfully different language.

Three details in the consumer wording repay attention. Anthropic does not merely disclaim ownership — it assigns to you whatever rights it may hold, which is the stronger position for a user. That assignment is conditional on your compliance with the Terms: breach them and its basis is in question. And the phrase “if any” is doing real work — Anthropic is not asserting that copyright exists in model outputs, only transferring whatever rights might exist. That qualifier is the honest heart of this topic, and we return to it below.

Anthropic's help centre is explicit about the practical scope: outputs “can also be integrated into your applications to power features within your products, generate content for your customers, analyze and structure your data, or improve internal workflows and productivity”. That covers essentially every commercial use people ask about — a Free-plan user selling a report they drafted in claude.ai chat is within the terms, as is a startup shipping a feature built on Sonnet 5 through the Claude API. The plan you are on changes what Claude can do, never whether you may sell the result.

#What are you not allowed to do?

The restrictions are narrow but firm.

• Do not use outputs to train competing models. Anthropic states: “Our Terms do not allow the use of Outputs to train models that are competitive with Anthropic's own.” Specifically named are general-purpose chatbots, models designed for open-ended text generation, using outputs as training targets, and reverse-engineering training methods. More broadly: “We prohibit customers from using our services to train or develop AI models without our written permission.”
• Do not build a competing service. The Consumer Terms prohibit using the Services to “develop any products or services that compete with our Services”.
• The Usage Policy applies to everything, separately from the Terms, regardless of plan or output ownership.
• You must be 18, or the local minimum age to consent, whichever is higher.

#Is your data used to train Claude?

The answer differs completely between the two contractual regimes, and conflating them is the most common error on this subject.

#Commercial plans: no, by default

Under the Commercial Terms — covering Team, Enterprise, the API, and Claude for Work, Government and Education — “Anthropic may not train models on Customer Content from Services”. Anthropic's privacy centre restates it plainly: by default, inputs and outputs from commercial products are not used to train its models.

One carve-out: if a user explicitly submits feedback, such as pressing the thumbs-up or thumbs-down button, that conversation may be used for training and is retained in secured back-end storage for up to five years. Team and Enterprise owners can turn this off entirely with the “Rate chats” setting under Data and Privacy. Worth doing on day one if your users handle anything sensitive.

#Consumer plans: opt-in since August 2025

Anthropic changed the consumer position on 28 August 2025, with existing users required to choose by 8 October 2025. The current rule is that Anthropic will use your chats and coding sessions to improve its models if you choose to allow it.

The policy applies to new or resumed chats and coding sessions, not retroactively, and it covers Free, Pro and Max including Claude Code. You can change the selection at any time in Privacy Settings, though opting out stops future use while training runs already under way continue. Safety-flagged conversations may still be analysed to improve detection. What each plan includes is set out on the free plan and pricing pages.

#Can you copyright what Claude writes?

Here the honest answer is uncomfortable. Anthropic can only give you whatever rights it holds; it cannot create copyright that does not exist. That is precisely what “if any” is doing in the Consumer Terms.

The status of purely machine-generated work is unsettled and jurisdiction-dependent. Several copyright offices, including in the United States, hold that copyright protects human authorship, and that material generated without meaningful human creative input may not qualify; other jurisdictions treat computer-generated works differently. The practical consequence: a paragraph you prompted for and pasted unchanged sits in weaker territory than the same paragraph after you have restructured, edited and integrated it into a work you authored.

• You can still use and sell it. Enforceable copyright and permitted commercial use are separate questions. Anthropic's terms settle the second; only a court settles the first.
• Human contribution matters. Selection, arrangement, editing and integration are the parts most likely to attract protection. Working iteratively, as described on Claude as a writing assistant and in the worked examples, produces a stronger position than one-shot generation.
• Ask about indemnity. Anthropic has published expanded IP indemnification for commercial and API customers. If third-party infringement risk is live for you, read the current terms rather than a summary.

#Do you have to disclose that you used Claude?

Anthropic's terms do not require it. Other things might. Client contracts increasingly contain AI clauses, academic and journalistic bodies set their own rules, some marketplaces require labelling of AI-assisted content, and regulators in several jurisdictions are moving on transparency obligations. None of that comes from Anthropic — it comes from whoever you are delivering to.

Two mechanical points. Anthropic documents that Claude marks AI-generated content on some surfaces, so a decision to stay quiet may be made for you by metadata you did not add. And AI-detection tools are unreliable in both directions, so “nobody will know” is a poor basis for a policy. The norm most teams land on: disclose to clients as a matter of process, do not annotate every internal document, and never let the question be decided ad hoc by whoever is under deadline.

#A compliance checklist before you ship

1. Confirm which terms govern you. Free, Pro and Max are Consumer Terms; Team, Enterprise and API are Commercial Terms — different documents, different obligations.
2. Set the training preference deliberately. On consumer plans, choose in Privacy Settings and tell people what you chose. On Team or Enterprise, consider disabling “Rate chats”.
3. Check nothing you build trains a model on outputs. Including small internal models, anything a vendor builds for you, and any automated pipeline that quietly accumulates outputs as a dataset.
4. Read the Usage Policy against your actual use case, not a generic description of it.
5. Decide your disclosure position and write it down before a client asks.
6. Verify facts in anything customer-facing. Owning an output says nothing about whether it is correct, and the liability for publishing something wrong is yours.
7. Keep editing records for assets whose ownership may need defending.
8. Re-check when you change plans. Moving from Pro to Team changes which contract applies.

#Frequently asked questions