Reviewers may use generative AI as an assistive tool in reviewing, subject to three requirements.
- Responsibility. AI may assist a reviewer but must not substitute for the reviewer’s own reading, assessment, and participation in the review process. Reviewers must personally evaluate the submission, exercise their own judgment, and participate in PC discussions as required. They remain fully responsible for every review, score, recommendation, and claim submitted in their name, and must independently verify AI-generated analysis on which they rely.
- Disclosure. Reviewers must disclose their use of generative AI in the designated field of the review form. The disclosure field will allow reviewers to indicate the nature and extent of the AI use. This information is visible to the Program Committee and is intended to provide transparency about the reviewing process and to help inform discussion of the review.
- Confidentiality. Confidential QIP material (such as the extended abstract, reviews, scores and PC discussions) may be provided only to an AI service that protects the confidentiality of that material. In particular, confidential material must not be used to train or improve the provider’s general models.