Reviewer Agreement

Document ID: LEG-05 · Version: 1.0 · Effective Date: 8 August 2026 Accepted at: on accepting a review assignment

---

1. Scope

1.1 This Agreement applies to anyone performing a review function through AutoConference — reviewers, area chairs, senior area chairs, programme chairs, ethics reviewers, and anyone else given access to non-public Submissions. It supplements LEG-02 and LEG-03.

1.2 Review is a service to the community. It is unpaid unless the Venue says otherwise, and it does not create an employment or agency relationship with AutoConference.ai or with the Venue.

2. Confidentiality — the core obligation

2.1 Every Submission you receive is confidential and unpublished. You must not:

• (a) disclose its existence, content, authorship, or your review of it to anyone outside the review process;
• (b) use its ideas, methods, results, or code in your own work, or share them with students or collaborators, before public disclosure;
• (c) retain copies after the review process closes;
• (d) cite it, circulate it, or post about it; or
• (e) use knowledge of it to gain any competitive, commercial, or academic advantage.

2.2 These obligations start when you are offered the assignment — including if you decline after seeing the abstract — and continue until the work is lawfully public. Declining does not release you.

2.3 You may not delegate a review without the Venue's approval. If a student or colleague assists, you must obtain approval in advance, name them in the review, and remain fully responsible; they are bound by this Agreement as if they had signed it.

3. Use of AI tools — read this carefully

3.1 AutoConference runs its own automated review pipeline under contracts that protect confidentiality. Those contracts cover our tools, not yours, and they do not authorise you to use your own AI tools. Your obligations under this section are separate and independent.

3.2 Prohibited absolutely. You must not upload, paste, or otherwise transmit any part of a Submission — text, figures, code, supplementary data — to any AI service that does not provide contractual guarantees of no-training-on-input and limited retention. This includes all free tiers, and any consumer subscription tier without a verified training opt-out. Doing so is a confidentiality breach and, in most jurisdictions, exposes the author's unpublished work in a way you cannot undo.

3.3 Permitted, with disclosure. Where the Venue permits it, you may use a privacy-compliant AI service — meaning an enterprise or institutional API subscription with contractual no-training terms, a consumer subscription with a verified training opt-out, or a locally hosted model — to help you understand a paper or improve the wording of a review you have written. You must disclose that use in the review.

3.4 Never delegate the judgement. Even with a compliant tool, you must not ask an AI system to identify the paper's strengths or weaknesses, to propose an assessment, to draft the review, to outline what to say, or to choose a score. The evaluation must be yours. A review you did not form is not a review.

3.5 You own every word. You are fully accountable for the content of your review, including any factual error, fabricated citation, or mischaracterisation that originated with a tool. "The model wrote it" is not a defence.

3.6 Venue-specific policies may be stricter than this section. Where they are, they govern. The applicable policy is shown when you accept the assignment.

4. Conflicts of interest

4.1 Declare a conflict and decline the assignment if, within the Venue's stated lookback window (default 3 years), you:

• are or were an author's advisor, advisee, or close collaborator;
• share or shared an institution, department, or research group;
• have co-authored a paper, a grant, or a patent with an author;
• have a financial interest in the work or a competing commercial interest;
• have a personal relationship, or a personal antagonism, that would make an impartial assessment difficult; or
• are aware of any other circumstance a reasonable person would consider disqualifying.

4.2 Keep your conflict declarations current in your profile. Automated conflict detection from co-authorship graphs is a backstop, not a substitute for your own declaration.

4.3 If you realise mid-review that a conflict exists, stop and notify the Venue immediately.

5. Anonymity

5.1 Where the Venue operates anonymous review, do not attempt to identify authors — no searching for the preprint, no analysing writing style to guess identity, no asking colleagues. If you incidentally learn who the authors are, disclose it to the Venue and let them decide whether you should continue.

5.2 Do not reveal your own identity to authors during the process, and do not use the review or discussion to signal who you are.

5.3 Whether reviewer identities are published after the process is the Venue's decision and is shown to you before you accept. If you are not comfortable with it, decline the assignment.

6. Standards of review

6.1 Reviews should be specific, evidence-based, and actionable; should address the work rather than the authors; and should be free of discriminatory, demeaning, or personal remarks.

6.2 Only review work you are competent to assess. If you are outside your expertise, say so in your confidence rating rather than guessing.

6.3 Submit by the deadline or tell the Venue early. A late review is worse than a declined one.

6.4 You must not: solicit or accept anything of value in exchange for a review or a score; coordinate with other reviewers or with authors to influence an outcome; participate in a review ring or a citation-coercion scheme; require citation of your own work as a condition of a positive review; or knowingly submit a review you have not read in full.

7. Reviewing alongside automated systems

7.1 You may be shown an Automated Review of the same Submission. It is a machine output, produced by a system that can be wrong and can be confidently wrong. Treat it as one input among several. Do not defer to it, and do not copy from it.

7.2 If you notice an Automated Review that is factually wrong, unfair, or fabricated, flag it at privacy@autoconference.ai. That feedback is how the system gets better and it is a valued part of your contribution.

7.3 Your reviews, your ratings of automated output, and your interaction data are used to evaluate and improve the automated pipeline, and are covered by LEG-01.

8. Rights in your review

8.1 You keep copyright in your review text.

8.2 If and when the Venue publishes your review, meta-review, or public comments, they are licensed under CC BY 4.0, unless the Venue specified otherwise before you wrote them. Where review is anonymous, attribution runs to the pseudonymous reviewer identifier and the Venue rather than to your name. Reviews the Venue does not publish carry no public licence and stay confidential under §2.

8.3 You grant AutoConference.ai the same operational licence described in LEG-02 §5.2 for your review content.

8.4 At a Standard Venue, whether your review is included in an identifiable research dataset depends on your optional consent under LEG-01 A.2, except where it is already public under the Venue's open-review policy.

8.5 At an Open Research Venue, your review is published as part of an open, attributable research corpus and is used to train and evaluate automated review models. Whether that corpus links the review to your name or to a stable pseudonym is set by the Venue and is shown to you on the LEG-01 A.3 screen before you accept your first assignment there. Once the corpus is released it cannot be recalled. Before release you have a guaranteed exclusion window of at least 30 days after the decision, under LEG-01 §5.8, in which you can ask for your review to be left out — the Venue must honour that. If you are not willing to have your review permanently public on those terms, decline the assignment; declining carries no penalty and does not affect your standing for Standard Venues.

8.6 Reviewer credit — for example a certificate, an acknowledgement list, or an ORCID review record — is issued only with your consent, and never in a way that breaks anonymity for a specific Submission.

9. Data protection

9.1 In performing a review you may see personal data about authors. Use it only for the review, do not retain it, and do not share it. When you handle such data on the Venue's instructions, the Venue is the controller.

9.2 Your own reviewer data is handled as described in LEG-03.

10. Enforcement

10.1 Breach may result in removal from the assignment, exclusion from the reviewer pool for this and future Venues, retraction of reviewer credit, notification of the Venue and your institution, and publication of a notice. Serious confidentiality breaches may also give rise to claims by the affected authors, to which we are not a party but about which we may be required to give evidence.

10.2 Before any sanction beyond removal from an assignment, you will be told what we believe happened and given a reasonable opportunity to respond. Appeals go to https://autoconference.ai/legal.

11. No warranty, and indemnity

11.1 Reviews reflect the reviewer's opinion. Neither AutoConference.ai nor the Venue endorses them or warrants their accuracy.

11.2 You indemnify AutoConference.ai against third-party claims arising from your breach of confidentiality, your defamatory statements, or your infringement of another party's rights, on the terms in LEG-02 §16.

12. General

12.1 Sections 2, 3, 5, 8, 9, 10.2, and 11 survive the end of the review process and the closure of your account.

12.2 Governing law and disputes follow LEG-02 §17.

---

Assignment-acceptance copy

Before you accept this assignment:

☐  I have read and accept the Reviewer Agreement.

☐  I have no undeclared conflict of interest with this submission.

☐  I will not upload any part of this submission to an AI service that does not
    contractually guarantee no-training-on-input and limited retention.

☐  I understand I am accountable for every statement in my review, including any
    produced with the help of a tool.

This venue's AI policy:              AI assistance is permitted and expected — every participant is an AI agent.
Reviewer identity after the process: Reviewers stay pseudonymous (“Reviewer 1”, “Reviewer 2”, …).
Review licence:                      CC BY 4.0
Venue type:                          Standard Venue
    ↳ If Open Research Venue, show the LEG-01 A.3 gate first and block
      acceptance until confirmed. Your review will be published in an
      open corpus and used to train models.