Terms of Service

These terms are an agreement between you and Rivul AI, which operates the service at rivul.ai. Creating an account or using the service means you accept them. If you do not accept them, do not use Rivul AI.

The Privacy Policy, the AI-Use Disclosure and the Copyright and Takedown page form part of these terms. Questions and notices go through the contact form at rivul.ai/contact.

Rivul AI is a web application for writing research papers. It provides an editor, a private library of PDFs you upload or fetch, a reader with highlight-to-cite, AI assistance for drafting and revising prose, retrieval and formatting of citation metadata from public scholarly databases, a set of review checks over your own draft, and export to LaTeX, Word and IEEE submission source archives.

Rivul AI is not a publisher, not a co-author, not a plagiarism checker, and not a substitute for reading the papers you cite. It does not verify that a source supports the claim you attached it to; the checks it runs are described in the product and in the AI-Use Disclosure in terms of what they actually measure.

Features and availability change as the product develops. We may add, alter or withdraw a feature. Where a change removes something you depend on we will try to give notice, and export is always available so that your work is never trapped by a change you did not want.

You must be at least sixteen, or older where your jurisdiction requires it, and legally able to enter this agreement. If you are using Rivul AI in the course of employment or study, you are confirming that you have whatever permission your institution requires.

Give accurate account details and keep the Google account you sign in with secure. You are responsible for everything done under your account, including by anyone you let use it. One person, one account: an account is not a shared departmental login, and sharing credentials is how one compromised password becomes everyone's problem.

Tell us through the contact form at rivul.ai/contact if you think someone else has access to your account. Securing that Google account, including changing its password, is the fastest thing you can do yourself.

Do not upload content you have no right to use. Rivul AI's core loop is researchers uploading published articles, and the rights in most of those belong to a publisher or an author rather than to you. Your institutional subscription generally lets you keep a copy for your own scholarly use; it does not let you upload a publisher's file on someone else's behalf, and it never covers a file you obtained from a source that had no right to give it to you. Prefer the open-access version when one exists. The Copyright and Takedown page explains how a rights holder gets a file removed and what happens to an account that accumulates notices.

Do not use Rivul AI to produce work that breaches an integrity policy that applies to you. Institutions, funders, journals and conferences each set their own rules on AI assistance, authorship and disclosure, those rules differ and they change, and complying with the ones that bind you is your responsibility rather than ours. The AI-Use Disclosure contains a statement you are free to adapt for a submission, but your venue's policy takes precedence over anything on this site.

Do not attempt to breach security, access another account, probe or circumvent rate limits and usage allowances, scrape the service, resell or sublicense access, use automated means to generate volume you would not generate by hand, or use Rivul AI to produce unlawful, defamatory or harassing material, or material designed to deceive about the provenance of research.

Do not use Rivul AI to fabricate research. This is stated separately from the rest because it is the one misuse that damages the people who rely on the literature rather than only us: generating a results section for an experiment that was never run, or citations chosen to look authoritative rather than because they were read, is outside what this service is for.

You own your content. The documents you write, the briefs, the references you assemble, your annotations and the files you upload remain yours, and nothing in these terms transfers any ownership to us. Rivul AI claims no authorship of, and no rights in, the papers you write with it.

You keep the copyright in AI-assisted text as between you and us. Whether machine-assisted output is copyrightable at all, and how a journal expects it to be disclosed, are questions of law and of editorial policy rather than of this contract, and both are moving. We make no representation about either.

To run the service we need a narrow licence, and it is narrow deliberately. You grant Rivul AI a non-exclusive, worldwide, royalty-free licence to host, store, copy, transmit, index, excerpt and display your content, and to send the parts of it that a request requires to the AI provider you connected and to the storage and infrastructure providers named in the Privacy Policy, for the sole purpose of operating the service for you. The licence lasts as long as you keep the content in Rivul AI and ends when you delete it or close your account, save for copies in backups until those age out.

That licence does not permit us to train a model on your content, to publish it, to share it with another user, to sell it, or to use it to develop anything other than the service as it operates for you. We do not do those things and this licence does not allow us to start.

This is the most important section on the page and the one with the most at stake for you, so it is written at length rather than compressed into a disclaimer.

AI output can be wrong. It can state something false with complete fluency, omit the qualification that made a finding defensible, mischaracterise a method, misattribute a result to the wrong study, or invert a direction of effect. This is a property of the technology and not a defect we expect to eliminate. Rivul AI reduces the specific risk of invented references by taking citation metadata from Crossref, DataCite, Semantic Scholar and arXiv records and formatting it by fixed rules rather than letting a model write the bibliography, and the review checks in the product report what they actually measured. None of that makes the prose correct.

You are the author. Every sentence you keep is yours, whether you typed it or accepted it. Before you submit anything for publication, or rely on it in a grant application, a thesis, a report or a clinical or engineering decision, you must read every AI-assisted passage, verify every claim against the underlying evidence, and open every citation and confirm that the source exists, says what you have said it says, and supports the specific claim you attached it to. Term overlap between your sentence and a cited passage is not support, and Rivul AI says so where it reports overlap.

Rivul AI is a drafting aid and not a source of truth, and it gives no warranty that any output is accurate, complete, current, original, non-infringing or fit for your purpose. To the fullest extent the law permits, we are not liable for the consequences of unverified output being published: that includes a correction, an erratum, a retraction, a rejected submission, a withdrawn or repaid grant, a finding of misconduct, a lost claim of priority, a disciplinary or professional consequence, or damage to your reputation or your co-authors'. Those consequences follow from a manuscript being submitted, and the decision to submit is yours alone.

If you are subject to an integrity policy that restricts or requires disclosure of AI assistance, follow it. If your co-authors have not agreed to the use of an AI tool on a shared manuscript, ask them before you use one. If you would not be comfortable telling a reviewer how a passage was produced, that is the signal to rewrite it.

Rivul AI depends on services operated by other companies, and the Privacy Policy names them and what each receives. The scholarly metadata databases, the hosting, storage, database, email and cache providers are independent third parties governed by their own terms.

AI features use an AI provider that you select in Settings and contract with directly, using your own API key. Your use of that provider is governed by your agreement with it, and it bills you directly. We are not responsible for that provider's output, availability, pricing, data practices, or any charges it makes to your key.

We are not responsible for the availability, output, pricing or practices of any of these third parties, and an outage or a change at one of them may degrade or interrupt Rivul AI. Where a provider's own terms constrain what you may do, those constraints reach you through this agreement: in particular, do not use AI output in a way your AI provider's policies prohibit.

Public research infrastructure such as Crossref, OpenAlex and arXiv is a shared resource maintained largely by non-profits. Rivul AI identifies itself and caches results to be a good citizen of it, and abusing it through Rivul AI is a breach of these terms.

When you connect an AI provider, you remain responsible for the API key you supply and for your account with that provider: for keeping the key confidential, for the limits and spending controls on it, and for every charge the provider makes to it. We strongly recommend that you set a spending limit with your provider before you connect a key, and that you use a key created only for Rivul AI.

Rivul AI stores one key per account, encrypted, and uses it only to send the AI requests you make while signed in. It is never shown back to your browser and is deleted when you remove it or delete your account. Saving a key for a different provider replaces and deletes the previous one.

Anyone who is signed in as you can use features that send requests on your key. Keep your Google account and your devices secure, and sign out on shared devices. If you believe your key has been exposed, revoke it with your provider immediately and remove it in Settings; revoking it with the provider is the step that stops further charges.

The same responsibilities apply to Rivul MCP and REST API keys you create: keep them confidential, revoke a key you no longer trust, and treat any client you connect as acting on your behalf.

To the fullest extent the law permits, we are not liable for charges made to your key, or for losses arising from its use or misuse, including use through your signed-in account or device, through a client you connected, or through a security incident at your provider.

Rivul AI is free. Every account can search, read, write, cite and export at no charge, and can store up to 5 PDFs at a time; files already stored above that number stay readable and deletable, and a new upload waits until the account is below it.

AI features run on an API key you supply, from an AI provider you choose. That provider charges you for your use under your own agreement with it. Rivul AI charges nothing for AI and pays for none of your model calls.

Rivul AI takes no payment and keeps no card on file, so there is nothing to renew, cancel or refund.

Rivul AI sells no AI allowance. Fair-use rate limits, counted per minute, per hour and per day and set well above a researcher's working pace, apply to AI requests, autocomplete and scholarly search to protect the service from automated traffic, and PDF storage is limited as described above. We may also pause a feature for everyone as an incident measure. Your AI provider may apply its own limits to your key.

Deliberately circumventing a limit, including by opening additional accounts to multiply one, is a breach of these terms.

You can stop using Rivul AI at any time, and you can delete your account from settings, yourself, without asking us. Deletion is immediate and permanent, so export your work first.

We may suspend or terminate an account that breaches these terms, that is being used unlawfully, that repeatedly attracts unrebutted copyright notices, or whose activity threatens the security or availability of the service for others. Where circumstances allow it we will warn you first and give you a chance to put it right, and we will describe what the problem is rather than closing an account in silence. Where the breach is serious or ongoing, suspension may be immediate.

If we terminate your account other than for a breach by you, we will give you at least thirty days' notice and an opportunity to export everything. On termination your licence to use Rivul AI ends and your content is deleted as described in the Privacy Policy.

We may discontinue Rivul AI entirely. If that happens we will give account holders at least sixty days' notice, and keep export working for the whole of that period.

Rivul AI is provided as it is and as it is available. To the fullest extent the law permits, we disclaim all warranties, express or implied, including any implied warranty of merchantability, fitness for a particular purpose, non-infringement, accuracy, and uninterrupted or error-free operation.

We do not warrant that the service will be available without interruption, that defects will be corrected, that it will be free of harmful code, that AI output will be accurate or original, that citation metadata retrieved from a third-party database will be correct, or that a retraction or preprint flag will catch every case. The retraction sweep depends on external registers being reachable and complete, and it is a help rather than a guarantee.

We protect the service and your data with reasonable technical and organisational measures, described in the Privacy Policy, including encryption of stored API keys. No method of transmission, storage or processing is completely secure, and we do not warrant that the service, your content or your keys can never be accessed, altered or disclosed without authorisation.

Nothing here excludes a liability that cannot lawfully be excluded. If you are a consumer, you have statutory rights that these terms do not affect, and where local consumer law gives you a stronger position than this section, that law wins.

To the fullest extent the law permits, neither party is liable to the other for indirect, incidental, special, consequential, exemplary or punitive damages, or for lost profits, lost revenue, lost data, lost publication or funding opportunities, or reputational harm, however caused and on any theory of liability, even if advised of the possibility.

To the fullest extent the law permits, our total liability for claims arising from these terms or the service is capped at one hundred United States dollars.

These limits do not apply to liability for death or personal injury caused by negligence, for fraud or fraudulent misrepresentation, or to any other liability that applicable law does not permit to be limited. They apply to the maximum extent permitted where a jurisdiction restricts them in part.

You accept that the allocation of risk in this section and in the AI output section is the basis on which Rivul AI is offered free of charge, and that verifying output before publication is the protection you have rather than a recovery from us afterwards.

You agree to indemnify and hold Rivul AI harmless against claims, losses and reasonable legal costs arising from content you uploaded that you had no right to upload, from your breach of these terms or of a law, and from your publication or other use of material produced with Rivul AI.

We will tell you promptly about any claim to which this applies, let you control the defence of it if you wish, and not settle it in a way that imposes an obligation on you without your consent.

We may update these terms. The current version is always on this page with its last-updated date at the top.

For a change that materially reduces your rights, increases your obligations, or widens what we may do with your content, we will email account holders at least thirty days before it takes effect. If you do not accept the change, you can export your work and delete your account before it takes effect. For a clarification, a correction, or a change required by law, the updated date is the notice.

Continuing to use Rivul AI after a change takes effect means you accept it. If you do not accept it, stop using the service and, if you wish, delete your account.

Nothing in this section removes a right you have to bring proceedings in your own country of residence where consumer law gives you that right, or to complain to a regulator.

Before formal proceedings, we ask that you describe the problem through the contact form at rivul.ai/contact. Early direct contact is usually the fastest and least expensive step for both sides.

If a provision of these terms is unenforceable, it is limited or severed to the minimum extent necessary and the rest continues to apply. A failure to enforce a provision is not a waiver of it.

You may not assign this agreement without our consent. We may assign it as part of a merger, acquisition or sale of assets, on notice to you, and these terms travel with it unless you are told otherwise and given the chance to delete your account.

These terms, together with the Privacy Policy, the AI-Use Disclosure and the Copyright and Takedown page, are the entire agreement between us about Rivul AI, and replace any earlier version of them.

Notices to you go to the email address on your account. Notices to us go through the contact form at rivul.ai/contact.