Recommended Free Tools
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
It depends, and no single rule covers every case. Whether AI training on your data is lawful turns on what the material is, how it was obtained and copied, what the training is for, which country’s law applies, and, in some cases, what the trained model can reproduce. A company’s statement about its training sources does not by itself show that every use behind the model was lawful.
Below, the question is broken into the factors that decide the outcome, followed by the position in the United States, the United Kingdom and the European Union. The position is drawn from official documents published or dated between May 2025 and July 2026. Where a status may have moved since then, that is flagged.
Why “your data” is not one legal category
“Your data” can mean several different things, and each raises a different legal question.
Free tools Windows power users keep installed
One-click scans. No signup required.
- Copyrighted expression: articles, photos, code, illustrations, music or other creative works you made. The core question is whether copying them for training needs permission or falls under an exception.
- Personal data: information that identifies you or can be linked to you, such as profile details, messages or images of you. The core question is whether a company may process it for training under privacy law.
- Confidential information: trade secrets, non-public business records, or material covered by a confidentiality term in a contract. The governing rules are contractual and trade-secret rules, which the official sources reviewed here do not analyze.
- Other material: facts, public records and data that are neither creative expression nor personal. Its treatment depends on the jurisdiction and facts, and it is outside the scope of this overview.
The factors that decide whether a use is lawful
Lawyers and regulators tend to work through the same questions. Going through them does not produce an answer on its own, but it shows which facts matter most.
#1 Best Overall
- Type of material. Is it copyrighted, personal, confidential or something else?
- Authorization and access. Was the material licensed, published on terms that allow the use, or lawfully accessed? Being able to see something is not the same as having permission to copy it.
- Copies made. Copies can be made when data is collected, when it is curated and when it is used in training. Each copying step can raise its own question.
- Purpose and method. Commercial use, how transformative the training is, and the technical method all feed into the analysis.
- Market effects. Whether the use substitutes for the original work or harms the market for it or for a license to it.
- Rights reservations and licenses. Whether the owner has reserved rights or set terms that restrict training.
- Model behavior. Whether the trained model memorizes material or reproduces it in outputs.
- Personal data. Whether identifiable people’s information is involved, and which privacy regime applies.
- Jurisdiction. Where the copying took place and which country’s law governs it.
These are issue-spotting questions, not a substitute for applying the law to a specific set of facts.
Does “publicly available” mean AI can use it?
Not on the terms the official sources set out. Public accessibility is one fact among many. The sources treat whether material was authorized or lawfully accessed as a separate question from whether a particular use is permitted. A work anyone can read online is still a copyrighted work unless a license, an exception or another basis applies. In the European Union, the owner may also have reserved text-and-data-mining rights under Article 4(3) of Directive (EU) 2019/790, which is why checking for a rights reservation matters even for material that is freely visible. The sources reviewed do not establish a general rule that public material may be used for training.
Can AI train on copyrighted work without permission?
Sometimes, depending on where the training happens and the facts of the copying. None of the three jurisdictions covered here has settled the question in a way that makes all training lawful or all unlicensed training unlawful.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
| Jurisdiction | Position in the official sources | Main open point | Source and date |
|---|---|---|---|
| United States | Fair use is decided case by case. The Copyright Office’s framework does not decide individual lawsuits. | Ongoing court decisions on fair use in AI training. | U.S. Copyright Office, Part 3 report, pre-publication version, May 2025 |
| United Kingdom | Existing copyright law remains the status quo while the government considers options. Reform is on hold until the government is confident it will meet its objectives. | The first UK court ruling on AI training was under appeal when the report was published. Not stated in the report: whether the appeal has since been decided. | UK Government, Report on Copyright and Artificial Intelligence, March 2026 |
| European Union | In-scope general-purpose AI model providers must have a copyright-compliance policy and publish a training-content summary. These are compliance and transparency duties, not a general license. | Not stated in the sources reviewed: how strictly the summaries will be assessed in practice. | Regulation (EU) 2024/1689 (AI Act), consolidated text dated 27 July 2026 |
United States
The U.S. Copyright Office’s Part 3 report, in its pre-publication form dated May 2025, analyzes fair use, licensing, acquisition, training and outputs. The Office says it is not opining on specific cases. It identifies four stages where copyright questions can arise: collection and curation of material, training itself, retrieval-augmented generation (where a system pulls in outside material when it answers), and outputs. The Office puts its approach this way:
“Without opining on specific cases, we provide an analytical framework for identifying relevant facts and policy considerations.”
The report says that more than 10,000 comments were received in response to the Office’s 2023 notice of inquiry on AI. It also notes that dozens of U.S. lawsuits were pending when the report was prepared. That lawsuit figure describes the situation in May 2025 and is not a current docket count.
Rank #3
The Office’s AI page says the Part 3 report was released in pre-publication form on May 9, 2025, and that a final version will be published later, without expected substantive changes to its analysis or conclusions. Check whether that final version has appeared before treating the pre-publication text as the Office’s final position.
United Kingdom
The UK government published its report on copyright and AI in March 2026 under sections 135 and 136 of the Data (Use and Access) Act 2025. It covers access to and use of copyright works in AI development, transparency, technical measures, licensing and enforcement. On reform, the government’s position is that it will wait until it is confident the changes will work:
“We will not introduce reforms to copyright law until we are confident that they will meet our objectives for the economy and UK citizens.”
The report says there is no consensus on how to balance rights holders’ reward and protection against developers’ access to quality content, and that a broad exception with an opt-out is no longer the government’s preferred route. This is a policy position, not a court ruling that settles all training conduct.
The report also discusses the first UK court ruling on copyright and AI training. Its interpretation was subject to appeal at the time, so treat that outcome as provisional until its current appellate status is checked. The report also explains that copyright is territorial: the place where copying happens can affect which copyright law governs, which matters when training involves copying carried out in more than one country.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesEuropean Union
The EU approach rests on binding duties, not on a general permission to train. Under Regulation (EU) 2024/1689, the AI Act, providers of in-scope general-purpose AI models must:
- put in place a policy to comply with EU copyright and related-rights law, including identifying and complying with rights reservations expressed under Article 4(3) of Directive (EU) 2019/790; and
- draw up and publish a sufficiently detailed summary of the training content, following the AI Office template, which the Commission published on 24 July 2025.
According to the UK government’s March 2026 report, which explains these provisions, the summaries are intended to be generally comprehensive and to help enforce copyright and other rights. They do not require a work-by-work assessment. A summary is a transparency duty that describes the training content a provider has disclosed. It does not show that every use was licensed or lawful. Whether a particular work may be used still turns on the copyright rules, any rights reservation, any license, the facts and any applicable exception.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Personal data is a separate question
When the material is personal data, copyright is often not the main issue. Privacy and data-protection law applies instead. The regimes that may be relevant include the GDPR in the EU, UK data-protection law, U.S. state privacy laws and others. Whether a particular person’s data may be used for a specific training purpose depends on the governing regime and the processing facts, including the purpose, the legal ground the provider relies on and where the data came from. The official sources reviewed do not resolve that question for any individual, and the copyright summaries above are not a substitute for a privacy analysis.
How to find out whether your data was used
This is the hardest part to answer. The official materials reviewed do not establish a general way for an individual to confirm whether a named AI system was trained on their work or information. The EU training-content summary is the most concrete public disclosure in these sources, but it applies only to in-scope providers and does not require work-by-work detail. In practice, you can:
- Find the provider’s published training-content summary, if it has one for the model you are concerned about, and check whether it describes the kinds of sources that matter to you.
- Read the provider’s privacy policy and terms for statements about training, and note any opt-out or objection mechanism it describes.
- Put specific questions to the provider in writing: which sources were used to train the model, whether your material or data is included, and how opt-out or objection requests are handled.
- If the material is personal data, use the access, objection or deletion rights your privacy law provides, and keep a record of each request and response.
A published opt-out or a written reply does not show that the request was honored in practice. The sources reviewed do not establish whether any particular provider honors opt-outs.
When to get qualified advice
This overview is not legal advice. For a dispute, a licensing negotiation or a compliance decision, a qualified lawyer should review:
Quick Recap
- the country or countries where the material was copied, and where the model is trained or deployed;
- the type of material and who owns or controls the rights, or whose personal data is involved;
- how the material was obtained, and whether any license, terms or rights reservation applied;
- which provider and model are involved, and what training or deployment details are known.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
