HotCRP, the widely-used conference management system for academic paper reviews, has published guidance addressing the use of AI agents and bot accounts on its platform. The announcement comes as researchers and reviewers increasingly experiment with large language models to assist in manuscript preparation, review summarization, and correspondence—raising thorny questions about authorship, integrity, and appropriate conduct in peer review.

What Is HotCRP?

For those unfamiliar, HotCRP (originally standing for something like 'HOT Conference Review Program') is a web-based system developed by Eddie Kohler at Harvard that handles paper submissions, reviewing assignments, discussion forums, and decision-making for academic conferences. Major venues in computer science—systems, programming languages, security—rely on it extensively. The platform's flexibility means conference organizers have significant control over policies governing how reviewers and authors interact with their infrastructure.

The AI Agent Problem

The core issue is straightforward: when a reviewer delegates their work to an autonomous agent, or when an author uses bots to craft submissions or respond to reviews, traditional notions of human authorship dissolve. HotCRP's latest guidance appears aimed at giving conference chairs explicit tools and policies to address this gray area—distinguishing between legitimate tool use (spell-checkers, reference managers) and delegating substantive intellectual work to systems that operate without direct oversight.

Why This Matters for the Community

The academic peer review system has always relied on implicit norms about reviewer independence and author honesty. AI agents shatter those assumptions. A reviewer could have a model read every submitted paper, generate thoughtful critiques, and submit them under their name—without ever reading a word themselves. Authors could use models to optimize responses to reviewer comments or even draft rebuttal letters in different 'writing styles.' The integrity of the review process depends on human judgment, and these tools threaten to commoditize that judgment.

Key Takeaways

  • HotCRP's policy update addresses AI agents directly, providing conference organizers with clearer governance options
  • The platform is used by major CS conferences for submissions and peer review management
  • Academic communities are still developing consensus on acceptable AI assistance in research workflows
  • Conference chairs gain more explicit controls over what automated systems can and cannot do on their venues' pages

The Bottom Line

This isn't just a technical tweak—it's a signal that the academic infrastructure community is waking up to the reality that autonomous agents will probe every system they can access. HotCRP's move sets a precedent for how peer review platforms handle the coming wave of AI-mediated submissions and reviews.