OSPAIC AI Usage Policy

This policy is current as of October 23, 2026. Any changes will be finalised by December 1 2026. It will be reviewed mid-2027.

Overview

One of the central tenets of OSPAIC is that advances in AI could present tremendous opportunities for humanity across every sphere of human activity, but that the unthinking deployment of AI to address any given problem is almost always a bad idea. AI should be used for some things but not for others; discerning which is which is among the central tasks before philosophers today. In particular, we must be alert to the negative externalities that AI use can create. We should aim to develop robust professional norms against uses of AI resulting in unacceptable externalities, reflected in clear policies and enforcement mechanisms.

The use of AI by authors raises three distinct concerns:

  1. AI writing, which OSPAIC prohibits outright (Section I);
  2. Authorship, the requirement that the named authors are genuinely responsible and accountable for the paper’s content (Section II);
  3. Misrepresentation, the requirement that readers and institutions receive an accurate picture of what the named authors contributed to the paper (Section III).

The policy additionally sets rules for reviewers’ use of AI (Section IV) and for the confidentiality of submissions and reviews in any AI processing undertaken by OSPAIC (Section V).

I. OSPAIC Prohibits AI Writing

Submissions to OSPAIC must be written by humans, for the following reasons.

We allow that certain uses of AI in academic writing might in themselves be relatively innocent. However, accommodating these innocent uses by permitting the submission of AI-authored papers opens the floodgates to an avalanche of bad AI writing (“slop”). Sifting through that avalanche to find the diamond in the slop presents untenable challenges. Requiring human authorship allows us to use AI detection software to prevent irresponsible researchers from DDoSing peer review. Writing a paper without relying on AI assistance also serves as a costly signal by the author; if the author doesn’t believe the paper is worth writing, the editors are justified in declining to find it worth reading. Use of AI for light copy-editing is permitted; for example correcting spelling and grammar, and reducing redundancy. We advise against sentence- or paragraph-level rewrites; see enforcement, below.

We will revisit this policy as AI capabilities advance, but we note that its justification does not depend on AI’s philosophical capabilities remaining static. Even if AI systems became superhuman at philosophy, the above points would remain.

II. Authorship

The prohibition on AI writing is necessary but not sufficient. A paper typed entirely by a human may still fail as a submission if its named authors are not responsible or accountable for its content.

Philosophical papers alone are not the sole (or perhaps even primary) unit of philosophical progress. OSPAIC is also a conference, premised on the value of bringing people together to exchange and build on one another’s ideas, and jointly to decide which questions deserve our collective attention. Authors submitting work that they did not author are less likely to be able to contribute to this kind of exchange.

Accordingly, papers may not be submitted to OSPAIC if the upstream contribution of AI to the paper is such that, had it come from a human colleague, that colleague would have a plausible claim to co-authorship. To illustrate, this means that AI systems are permitted to contribute upstream in ways similar to a research assistant, someone giving comments at a conference, reviewers for a journal, or an instructor teaching the author about some relevant technical subject matter, or in other ways that would not justify an authorial role.

III. Misrepresentation and Disclosure

Professional journals also have an institutional role, providing evidence in professional evaluation (e.g. in hiring, tenure, and promotion). With this in view, authors must submit a statement describing any AI methodology involved in writing the paper, focusing on the kinds of contributions that are relevant to the counterfactual test described above (authors who use AI agents to file reimbursement claims that free up their time to write do not need to declare this fact as part of their methodology).

These statements of AI methodology will be part of the published record; false statements are research integrity violations.

IV. Reviewers and use of AI by OSPAIC

Reviewers must likewise author their reviews, and are not permitted to use LLMs for any purpose in doing so.

The OSPAIC editorial team will use AI in the following ways: for logistical purposes; to identify AI-written papers; to screen out very low-quality papers; and to provide a technical review to papers invited for submission to the journal. By "technical review" we mean an assessment of the correctness of the claims made in the paper, especially logical validity and the truth of any verifiable empirical claims, as distinct from an assessment of the paper's significance or novelty. Papers will not be rejected on the basis of this technical review.

V. Privacy of Submitted Research and Reviews

Any processing of submissions or reviews by OSPAIC will be undertaken through AI tools covered by licenses whereby there is (a) no use of inputs for model training, and (b) zero or minimal data retention. Public chat interfaces may not be used by editors, reviewers, or staff for any material containing submission content, in whole or in part. Where feasible, OSPAIC will prefer self-hosted or locally-run models, which eliminate third-party data exposure.

VI. Enforcement

All papers will be screened by the best available AI detection software both at submission and as new models are released throughout the paper’s lifecycle, including after publication. Any submitted paper that is discovered to have been substantially AI-written will be rejected or retracted. Authors who have more than one paper rejected by the initial screening process will be banned from submitting further papers that year. If papers are identified as having been AI-written subsequent to the initial screen, the author will be banned from submitting to OSPAIC for the following year. If identified at the in-person conference or after, the author may be banned from submitting to OSPAIC indefinitely.

The perfect AI detector would be able to identify the precise proportion of tokens in a given essay that were generated by AI, as opposed to by a human. If we had such a detector, we would treat greater than 10% AI use as disqualifying, in order to permit copy-editing. Lacking a perfect oracle, we will use available AI writing detectors to approximate it as best we can.

Authorship and misrepresentation are challenging to enforce, and substantially rely on researchers’ integrity and mutual enforcement. All papers will be presented at an in-person conference; authors’ ability to defend and expand on the claims in their submissions will factor into the decision to invite submissions to the annual volume. Authors discovered to have misrepresented their contribution will be subject to rejection and retraction and the same schedule of bans as with AI writing.

AI and Translation

OSPAIC is an English-language journal and conference. As such, papers must be written in English, and authors need to be able to present and defend their paper at the conference in English. So there is no exception to our policy for the use of AI to translate papers written in other languages.