A new framework called Symposium aims to improve trust among researchers by creating auditable records of AI agents' activities. This involves recording analyses, hypotheses, data, and scientific discourse in a shared, immutable history that can be used for trust assessments. The framework captures structured claims, evidence citations, assumptions, and explicit declarations of what material is or isn't considered evidence. A working implementation is provided to enable users to set up their own Symposium community. --- Why it matters: This matters because it addresses the issue of trust in AI-driven research communities, where diverse systems operate in a rapidly evolving environment. By providing a shared record of activities and decisions, researchers can build on prior work and make informed trust assessments. Source: https://arxiv.org/abs/2608.19511