AI Attribution Toolkit (aiattribution.github.io)

IBM Research released an AI Attribution Toolkit — a practical attribution statement format and assets (copyable text, SVG, HTML) designed to give granular, standardized disclosure about how AI contributed to content. Built from CHI 2025 research, the statement comes in full and abbreviated forms (with optional icons) and encodes fields such as an identifier, version number, whether a human reviewed the output, what the AI was used for, and level/type of contribution. Statements can be linked to the toolkit website for context and downloaded for direct embedding in digital content, letting creators assert anything from “no AI used” to “AI generated majority of content” with precise metadata. For the AI/ML community this matters because generic labels like “assisted by AI” no longer suffice: researchers, platform builders, publishers and auditors need consistent, machine- and human-interpretable signals about provenance, ownership and trustworthiness. The toolkit’s structured approach supports better attribution practices, reproducibility, moderation, and regulatory compliance, and can be adopted into publishing workflows and metadata standards. IBM frames the assets as open-source and solicits feedback via GitHub and a feedback form to refine formats and adoption pathways.