Acceptable Use Policy

#1. Scope and nature of this document

This Acceptable Use Policy sets out how the Service may be used and forms an integral part of the agreement for the use of the Service. Capitalised terms — in particular Operator, Client, Consumer, Account, Service, Source, Goal, Cluster, Content and Source Material — have the meaning given to them in the “Definitions” section of the Terms of Service.

The document falls into two parts of a different character. The sections from “Configuring Sources” to “Prohibited uses” set out the Client’s obligations and cover only what the Client actually controls: the configuration of its Account, Goals and Sources, the instructions it enters, and the decision to publish the Content produced; the section “Consequences of a breach” describes what follows from a breach of those obligations. The section “How the fetching engine behaves” imposes no obligation on the Client — it contains the Operator’s statements.

That division is deliberate. Material is fetched from Sources by the Operator’s infrastructure, not by the Client. An obligation about the way that infrastructure works, if imposed on the Client, would be an empty one, because the Client has no means by which to perform it. The Operator therefore describes the behaviour of its own engine as a statement for whose truth it answers itself.

This Policy applies to every Client and every Account, whatever the plan, and to any person to whom the Client gives access to its Account or to the Content produced in it. The Client is responsible for the acts of such persons as for its own.

#2. Configuring Sources — the Client’s obligations

Which Sources the Service observes is decided by the Client alone, through the configuration of its Goals. By configuring a Source and keeping it in a Goal’s configuration, the Client represents that it is entitled to instruct the Operator to fetch and process material from that Source for the purposes of its Goals, and assumes the obligations set out below.

• Closed access. The Client shall not configure as a Source any service, or part of a service, access to which requires authentication, registration or a paid subscription that the Client does not hold. The prohibition extends to circumventing such a requirement by any means, in particular by pointing at an address that bypasses the login page, at a distribution channel intended for subscribers only, or at a copy of the content made available by a person who was not entitled to do so.
• No reproduction of a Source. The Client shall not configure a Goal in a way that results in the reproduction of a substantial part — whether qualitatively or quantitatively — of a single Source, in particular of its archive, of its entire body of publications, or of the whole of its current output fetched systematically. The Service exists to observe many Sources and to select material from them against the criteria of a Goal, not to build a copy of someone else’s service.
• The Source’s own terms. The Client shall not configure, or keep in its configuration, a Source whose terms of use exclude automated fetching of content for the purposes for which the Client uses the Service.
• Responding to takedown requests. Where the rightholder of an item of material, its publisher, or a person whose data the material contains requires that the use of that material cease or that it be deleted, the Client shall consider the request without undue delay and no later than within 7 days of receiving it. Where the request is well founded, the Client shall remove the material from its Clusters or remove the Source from the Goal’s configuration, and shall withdraw any Content published on the basis of that material to the extent the request concerns it.
• Passing requests on. Where giving effect to such a request calls for action on the Operator’s side — in particular the deletion of material beyond the reach of the Service panel — the Client shall forward the request to the Operator’s contact address without undue delay. Requests received by the Operator directly which concern Sources configured by the Client are passed on by the Operator to that Client.

The full text of the Client’s representations about its rights to the Sources it configures, their scope, and the terms on which the Operator is indemnified against third-party claims — together with the carve-out for Consumers and for natural persons contracting in a capacity that is not professional — is set out in the section Rights to content of the Terms of Service. This section neither replaces nor modifies those representations; it states only the obligations the Client performs in the course of using the Service.

#3. Publishing Content — the Client’s responsibility

The Client decides whether, where and in what form to publish Content produced in the Service, and answers for that publication as for material of its own. The Operator performs no editorial function, does not review Content before publication, and does not assess whether the publication of any particular Content is lawful.

The Client warrants that it holds the rights necessary to publish the Content to the extent that it publishes it — including rights to any graphics, photographs and quotations used in the Content — and that the publication infringes no third-party rights or personality rights and breaches neither press law nor the sector-specific rules applicable to its activity.

The fact that Content was produced from a Cluster does not settle whether it may be published. It is for the Client to ensure that published Content does not take over the creative elements of Source Material beyond the limits of permitted use, in particular beyond the right of quotation, and that it credits the source wherever the law or the standards of responsible journalism so require.

The Client shall review every item of Content editorially before publishing or otherwise using it, and shall in particular check it against the Source Material and verify the accuracy of personal data, dates, figures and quotations. Publishing Content automatically without such review is done at the Client’s own risk.

The Client’s warranty as to rights in Source Material, and its obligation to indemnify the Operator against third-party claims, are set out in the section Rights to content of the Terms of Service. The indemnity in its full scope does not bind a Consumer, nor a natural person contracting in a capacity that is not professional — as against those Clients, liability for a breach of the representations given is governed by the general rules of the Civil Code alone.

#4. Marking AI-generated Content

Content in the Service is produced with the involvement of artificial intelligence systems. In publishing Content the Client acts as the deployer of such a system and puts its output into public circulation; the resulting disclosure duties fall on the Client.

The Client undertakes to mark every published item of Content that has been generated, or materially reworked, by an artificial intelligence system — legibly, visibly and in a manner noticeable on ordinary consumption of the material, at the latest at the time of the first disclosure to recipients. A record in the metadata, a mention in a technical footer, or a notice that the recipient must take a further step to see does not satisfy this obligation.

This applies in particular to text published for the purpose of informing the public on matters of public interest. The Client shall not present Content as material produced by a human being alone where that is not the case.

The scope of the exception for Content that has undergone human editorial review and for which a natural or legal person holds editorial responsibility is described in the section AI-generated content of the Terms of Service. Whether that exception is met in a given case is for the Client, as the publisher, to assess; the Operator does not assess it on the Client’s behalf.

#5. Prohibited uses

The Client shall not use the Service for any of the following purposes, nor enable third parties to do so:

• Disinformation. Producing and disseminating Content that presents invented events, statements or figures as fact, and Content calculated to mislead its recipients as to matters of health, public safety, the conduct of an election or referendum, or the state of a market.
• Impersonation. Publishing Content that impersonates another person, newsroom, public authority or brand, including Content attributing statements to people who did not make them, and material suggesting authorship or endorsement by a party that has not endorsed it.
• Unlawful content. Producing or disseminating content that infringes copyright, related rights, the sui generis database right, personality rights, trade secrets or data protection law, and content inciting violence or hatred, insulting a section of the population, promoting totalitarian regimes, or amounting to the harassment of an identified person.
• Manipulation through scale. Using the Service for automated publication calculated to manipulate search results, public opinion or market prices, including the operation of multiple outlets presented as independent of one another.
• Circumventing the limits of the Service. Opening multiple Accounts in order to evade the limits of a plan, loading the Operator’s infrastructure beyond what ordinary use entails, and reselling access to the Service or sharing an Account with third parties otherwise than as the Terms of Service allow.
• Circumventing model safeguards. Entering instructions designed to defeat the safeguards of models made available by third parties, and using the Service in a manner inconsistent with those models’ own terms of use.

The list above is illustrative in the sense that it does not displace the duty to use the Service lawfully and in good faith. It is not, however, a basis for extending these prohibitions to lawful uses that it does not name.

#6. Consequences of a breach

Where the Operator has reasonable grounds to suspect a breach of this Policy, it shall call on the Client to cease the breach and remedy its effects within a period of no less than 7 days. The Operator may act immediately, without a prior call, where the breach endangers the security of the Service, the data of other Clients or the rights of third parties, or where an obligation to act immediately follows from the law or from a decision of a competent authority.

The Operator may suspend access to the Account in whole or in part, take a designated Goal out of further processing, remove a Source from a configuration, or halt the delivery of Content to the Client’s channels. Any such measure is applied to an extent proportionate to the breach and for no longer than is necessary to bring it to an end; once the cause has ceased, the Operator restores access.

The Operator gives the Client notice of the measure applied at the email address associated with the Account, stating its cause, its extent and its expected duration. The Client may challenge the measure through the complaints procedure described in the section Complaints of the Terms of Service.

Where the breach is gross or is repeated despite a call to cease it, the Operator may terminate the agreement for the use of the Service in the manner and on the notice described in the section Termination of the Terms of Service. The application of the measures described in this section leaves the Operator’s further rights unaffected, in particular claims in damages on the terms set out in the section Liability of the Terms of Service.

Neither suspension nor termination for breach deprives the Client of the right to export its own data on the terms described in the Terms of Service, unless the export is legally impossible or would itself constitute a further breach.

#7. How the fetching engine behaves — the Operator’s statements

The provisions below impose no obligation on the Client. They are the Operator’s statements about how the engine by which the Service fetches material from configured Sources behaves, and they describe the position as at the date this Policy takes effect. A material change in that behaviour requires this document to be amended in the manner applicable to amendments of the Terms of Service. The Operator states as follows:

• We fetch publicly available pages only — pages the Source’s server serves without authentication. A fetch consists of sending an ordinary request for the document at the given address and reading the server’s response.
• We do not attempt to circumvent paywalls, subscription walls or login mechanisms.
• We use no credentials — no logins, passwords, access tokens or session cookies — to obtain access to paid or otherwise closed content. The fetching engine has no technical means of authenticating itself to a Source.
• Paywall blocks we detect are removed from the fetched content: where the text contains a demand to purchase access or to log in, the remainder of the text is cut off and does not pass into further processing. We do not warrant that detection covers every form such a demand may take.
• We do not train or fine-tune machine learning models on fetched Source Material or on the Client’s Content. Material and Content are passed to models solely as the input of a single call, for the purpose of performing the Service for the Client.

The language models to which the Service directs its calls are made available by third parties. They are listed, together with their processing locations and transfer bases, in the List of sub-processors.