TourMaG.com – What are the main legal risks for a tourism professional who uses AI to create content?
Me Chloé Rezlan: I see three main ones. The first is un intentional plagiarism. A generative AI outputs results by drawing on a vast amount of existing content, much of which is protected. Texts, photos, or videos generated can thus closely reproduce a pre-existing work without the user realizing it. The publisher then exposes themselves to an infringement action.
The second is the lack of protection. Many professionals think that the generated content belongs to them, like content produced by a communications agency. That is not the case. A content piece produced largely by a machine is not protected by copyright.
The third is the lack of transparency. The European AI Regulation, the AI Act, now imposes transparency obligations, particularly for content that imitates reality (hyperreal manipulation). Ignoring them puts you in a non-compliant position. There is also a risk that the content could be deemed as misleading commercial practices if it could deceive the consumer.
More broadly, the danger lies in using AI without a framework. In many companies, every employee selects their tool, their method, and what they transmit. That is why I recommend two things: train the teams and adopt an AI usage charter.
Me Chloé Rezlan: I see three main ones. The first is un intentional plagiarism. A generative AI outputs results by drawing on a vast amount of existing content, much of which is protected. Texts, photos, or videos generated can thus closely reproduce a pre-existing work without the user realizing it. The publisher then exposes themselves to an infringement action.
The second is the lack of protection. Many professionals think that the generated content belongs to them, like content produced by a communications agency. That is not the case. A content piece produced largely by a machine is not protected by copyright.
The third is the lack of transparency. The European AI Regulation, the AI Act, now imposes transparency obligations, particularly for content that imitates reality (hyperreal manipulation). Ignoring them puts you in a non-compliant position. There is also a risk that the content could be deemed as misleading commercial practices if it could deceive the consumer.
More broadly, the danger lies in using AI without a framework. In many companies, every employee selects their tool, their method, and what they transmit. That is why I recommend two things: train the teams and adopt an AI usage charter.
“Only human creation can be protected by copyright”
TourMaG.com – Who is responsible when content generated by AI contains an error, misleading information, or infringes the rights of a third party?Me Chloé Rezlan: The professional who publishes. It is not the provider of the LLM (i.e., large language model). The terms of use of generative AI solutions almost always exclude their liability for the results produced.
The Air Canada case illustrates this well. In 2024, a Canadian court found the company liable because its chatbot gave a passenger incorrect information about a bereavement-related discounted fare. Air Canada argued that the chatbot was responsible for its own responses. The court rejected the argument: the company is responsible for all information published on its site, whether it comes from an employee, a static page, or an AI bot.
In tourism, the stakes are particularly high. The Tourism Code requires precise pre-contractual information to the traveler, and the Consumer Code sanctions misleading commercial practices. An AI-created hotel description, an approximate beach distance, or an image that embellishes reality can expose the professional to liability. It is the professional who answers to the client, not the algorithm.
TourMaG.com – Copyright, right of publicity, trademarks: what precautions should be taken before publishing content created or modified by AI?
Me Chloé Rezlan: The first precaution is taken when drafting the prompt. Avoid asking the tool to draw too directly from protected works. Asking for a poster “in the style of” a given illustrator or a text “in the manner of” a given author greatly increases the risk of infringement. It is better to describe the ambience, colors, and emotion you want than to cite a work or a creator.
The same vigilance applies to rights of publicity and trademarks. A generated visual that resembles a real person, especially a public figure, or that shows a logo or a distinctive brand element, should not be published as is.
Then, to have a chance at protecting one’s content, keep in mind a simple principle: only human creation can be protected by copyright. In French law, a work must bear the imprint of the author’s personality, and a machine has no personality.
To claim protection, one must therefore demonstrate a genuine human creative contribution. Concretely, this means reworking the result multiple times, making choices, modifying, composing.
One must also document this process by keeping the successive versions, the prompts, and the modifications made. In case of dispute, this trace will allow proving human intervention.
“Do not transmit personal data, nor sensitive data, nor confidential information”
TourMaG.com – What data can a tourism professional transmit to an AI and which must be protected at all costs?
Me Chloé Rezlan: The basic rule is to transmit neither personal data, nor sensitive data, nor confidential information.
Tourism professionals handle a huge amount of personal data: travelers’ identities, passport numbers, contact details, booking histories. Some are even GDPR-sensitive data, such as health information (allergies, mobility limitations, diet related to a medical condition). Copying a client file into a consumer-grade AI tool to draft a personalized program potentially transfers these data to a third party, sometimes outside the European Union, and sometimes to train the model.
It is also necessary to protect the company’s strategic information: negotiated rate cards with providers, contracts, margins, client files, development projects.
In practice (ideally…), one should systematically anonymize, favor professional versions of tools that guarantee non-use of data, and verify where they are hosted.
TourMaG.com – What is the most common legal error you see among professionals who use AI to create content?
Me Chloé Rezlan: They use it as if there were no rules. AI arrived in companies through usages, at a breathtaking speed. Many professionals therefore still treat these tools as a free-for-all or decide not to worry about them.
Yet rules exist, and they are multiplying. The AI Act, for instance, imposes a transparency obligation for hyperreal content, i.e., images, sounds, or videos generated or manipulated that resemble real people, places, or events and may appear authentic.
In tourism, this is very concrete: a promotional video showing a paradise beach that does not exist in reality, or a fabricated customer testimonial filmed, must be clearly identified as AI-generated. Otherwise, in addition to lack of transparency, there is a risk of misleading commercial practices.
Me Chloé Rezlan: The basic rule is to transmit neither personal data, nor sensitive data, nor confidential information.
Tourism professionals handle a huge amount of personal data: travelers’ identities, passport numbers, contact details, booking histories. Some are even GDPR-sensitive data, such as health information (allergies, mobility limitations, diet related to a medical condition). Copying a client file into a consumer-grade AI tool to draft a personalized program potentially transfers these data to a third party, sometimes outside the European Union, and sometimes to train the model.
It is also necessary to protect the company’s strategic information: negotiated rate cards with providers, contracts, margins, client files, development projects.
In practice (ideally…), one should systematically anonymize, favor professional versions of tools that guarantee non-use of data, and verify where they are hosted.
TourMaG.com – What is the most common legal error you see among professionals who use AI to create content?
Me Chloé Rezlan: They use it as if there were no rules. AI arrived in companies through usages, at a breathtaking speed. Many professionals therefore still treat these tools as a free-for-all or decide not to worry about them.
Yet rules exist, and they are multiplying. The AI Act, for instance, imposes a transparency obligation for hyperreal content, i.e., images, sounds, or videos generated or manipulated that resemble real people, places, or events and may appear authentic.
In tourism, this is very concrete: a promotional video showing a paradise beach that does not exist in reality, or a fabricated customer testimonial filmed, must be clearly identified as AI-generated. Otherwise, in addition to lack of transparency, there is a risk of misleading commercial practices.
Five best practices for using AI
1. Treat AI as a real business project. It isn’t a tool that everyone installs on their own. That means identifying use cases, appointing a point person, training teams, and budgeting. Training is essential: an employee who understands the risks makes better daily choices.
2. Adopt an usage charter. It specifies what is allowed and what isn’t, which tools are approved, which data never to transmit, and the validation steps before publication. It’s a simple, practical document that secures the company and reassures teams. Note that: Entreprises du Voyage will soon publish a guide to help its members draft this charter.
3. Ensure the tools are appropriate. Terms of use are not all equal. Some free versions prohibit commercial use of generated content or reuse the input data. It’s important to ensure the tool permits commercial exploitation, guarantees confidentiality, and offers GDPR-compliant hosting.
4. Systematically revise the content. Generated content is a starting point, not a finished product. Rereading it, verifying facts, correcting, enriching, and adapting to your brand helps limit errors, reduce infringement risk, and improve chances of protection.
5. Document the creative process. Keep the prompts, the successive versions, and the changes made. This traceability serves both to prove human contribution for rights claims and to demonstrate good faith in case of dispute.
2. Adopt an usage charter. It specifies what is allowed and what isn’t, which tools are approved, which data never to transmit, and the validation steps before publication. It’s a simple, practical document that secures the company and reassures teams. Note that: Entreprises du Voyage will soon publish a guide to help its members draft this charter.
3. Ensure the tools are appropriate. Terms of use are not all equal. Some free versions prohibit commercial use of generated content or reuse the input data. It’s important to ensure the tool permits commercial exploitation, guarantees confidentiality, and offers GDPR-compliant hosting.
4. Systematically revise the content. Generated content is a starting point, not a finished product. Rereading it, verifying facts, correcting, enriching, and adapting to your brand helps limit errors, reduce infringement risk, and improve chances of protection.
5. Document the creative process. Keep the prompts, the successive versions, and the changes made. This traceability serves both to prove human contribution for rights claims and to demonstrate good faith in case of dispute.
