AITC 2026 — Author Guidelines

About the Conference

The AI Transparency Conference (AITC) 2026 welcomes extended abstract submissions presenting novel work at the intersection of AI transparency, agent behavior, and safety. Every submission will go through a double-blind peer review process. Authors of accepted work will present at the conference on June 5–6, 2026.

Topics of Interest

AITC 2026 is organized around three interconnected themes:

Detect — Agent Behaviour & Collaboration

Methods for observing, measuring, and characterizing the behavior of autonomous AI agents, including multi-agent interactions and emergent collaboration patterns.

Understand — Transparency & Interpretability

Techniques for making AI decision-making processes legible to humans — from mechanistic interpretability to explanation interfaces and audit tools.

Control — AI Control & AI Safety

Frameworks and mechanisms for maintaining meaningful human oversight of AI systems, including alignment verification, safety testing, and governance protocols.

We welcome contributions that span multiple themes or propose new perspectives on these challenges.

Important Dates

These deadlines are firm — we are unable to accommodate late submissions or extensions for any reason.

Submission Instructions

Where to Submit

All submissions are handled through OpenReview: https://openreview.net/group?id=AITC/2026/Conference

Every author listed on a submission must hold an active OpenReview account. We recommend signing up with an institutional email address, as this activates your account immediately. Non-institutional addresses may take up to two weeks to process.

Format and Length

AITC 2026 accepts extended abstracts only. The main text must not exceed 2 pages. References may use as many additional pages as needed. Both must be combined into a single PDF for upload.

An optional appendix may follow the references (e.g., for extra figures, proofs, or experimental details). However, reviewers are under no obligation to consult appendix material, so everything essential to evaluating your contribution must fit within the 2-page main body.

Submissions that exceed the 2-page main text limit will be automatically desk-rejected.

For the final camera-ready versions of accepted papers, the main body is typically between 8 and 12 pages long with a hard limit of 20 pages for the entire paper including references and appendices.

Template

All submissions must be prepared using the official AITC 2026 LaTeX template, which is based on the acmart document class (using the sigconf format with the nonacm option). The template is available on Overleaf:

https://www.overleaf.com/read/xpwwwwchcfdf#bebe68

To get started, open the link and click Menu → Copy Project to create your own editable copy.

The template is pre-configured for anonymous submission. Authors should not modify margins, font sizes, column layout, or spacing. Work prepared with other typesetting tools or that deviates from the required format will not be considered.

For the submission version, use the document class options as provided in the template:

\documentclass[sigconf, anonymous, review, nonacm]{acmart}

For the camera-ready version of accepted papers, remove anonymous and review:

\documentclass[sigconf, nonacm]{acmart}

Supporting Material

You may optionally upload code, data, or other artifacts that back up your findings. All supporting material must be fully anonymized — strip out author names, license headers, institutional paths, and anything else that could identify you. You can either attach a .zip archive or point to an anonymous repository.

Supporting material is due at the same time as the extended abstract — there is no separate deadline.

Review and Anonymity Policies

Double-Blind Review

AITC 2026 operates under a strict double-blind protocol: neither authors nor reviewers will know each other’s identities at any point during review.

To preserve anonymity in your submission:

• Do not include author names, affiliations, or funding details anywhere in the PDF.
• When referencing your own prior work, write about it as if someone else authored it (e.g., “Prior work by [Author] demonstrated…”).
• Do not link to personal websites, named GitHub repositories, or any other resource that could reveal your identity.

Work that compromises the double-blind process — whether intentionally or through oversight — may be removed from consideration.

Preprints (e.g., on arXiv) are fine. Their existence does not break anonymity as long as your AITC submission does not explicitly point to them in a way that identifies you as the author.

Prior and Parallel Publications

You may not submit work that has already been published or accepted at an archival venue (i.e., a conference or journal with formal proceedings). Submitting substantially the same work in parallel to another archival venue is also not permitted.

That said, earlier versions presented at workshops, symposia, or other non-archival venues do not conflict with this policy. Preprint server postings (arXiv, SSRN, etc.) are likewise permitted.

LLM Usage

Authors are free to use large language models as everyday productivity tools — for prose polishing, brainstorming, coding assistance, and similar tasks. No special disclosure is needed for this kind of routine use.

However, when an LLM has played a substantial role in shaping the research direction or generating significant portions of the text, authors must include a section titled “LLM Usage Disclosure” that clearly describes what the model contributed. This section can go in the appendix and is exempt from the 2-page limit.

Omitting this disclosure when it applies is grounds for desk rejection.

A few additional ground rules:

• Accountability stays with the authors. You are responsible for every claim in your submission, including anything an LLM helped produce. Fabricated results, hallucinated citations, or unattributed text generated by a model will be treated the same as any other form of misconduct.
• LLMs cannot be listed as authors.
• Prompt injection is prohibited. Do not embed hidden instructions aimed at influencing automated review tools. Submissions found to contain prompt injection will be desk-rejected immediately.

Ethics Statement (Optional)

If your work raises ethical questions — for instance, around human subjects, sensitive data, dual-use potential, fairness, or privacy — we encourage you to add a short Ethics Statement before the references. This is optional, does not count toward the page limit, and should be kept to roughly half a page.

Reproducibility

While not mandatory, we strongly encourage authors to take concrete steps toward making their results reproducible. This could mean submitting code, describing your experimental setup in detail, specifying hyperparameters, or providing links to anonymized resources. Reviewers are asked to consider reproducibility as part of their assessment.

After Acceptance

Camera-Ready Version

Authors of accepted work will be asked to upload a final camera-ready PDF through OpenReview by May 15, 2026 (23:59 CET). Detailed formatting instructions will be provided with the acceptance notification. The same LaTeX template and page limits apply.

Poster Presentation

All accepted extended abstracts will be presented as posters during the conference on June 5–6, 2026. Please prepare your poster in A0 portrait format (841 × 1189 mm). Specific scheduling and logistical details will follow closer to the event.

Code of Conduct

Everyone involved in AITC 2026 — authors, reviewers, organizers, and attendees — must follow the AITC Code of Conduct.

This applies to every aspect of the conference, from submission through to on-site participation.

Questions?

If anything in these guidelines is unclear or you have a question not covered here, reach out to us at contact@coairesearch.org.