Stability AI’s Annual Integrity Transparency Report

Stability AI released its annual integrity transparency report for April 2024 to April 2025.

The report details safety-by-design practices across video, image, 3D, and audio models, including dataset screening, NSFW classifiers, content filters, safety LoRAs, and red teaming.

Stability AI says it does not intentionally source data from the dark web, adult sites, or paywalled sources.

It also says it detected zero CSAM/CSEM in its training datasets during the period and reports detected CSAM to NCMEC.

The report highlights ongoing safeguards for its models and API, underscoring trust, compliance, and responsible AI oversight.