Stability AI

Ce qu'il fait

'Stable Safety' — the publicly branded safety function covering open-weight image and audio models, built around six principles including Safety-First Open Source and Banning AI Misuse, plus an annual Integrity Transparency Report.

Évaluation franche

No public accounts found of response rates to safety@stability.ai. Stability's safety posture has been heavily criticised in the context of open-weight image models and CSAM (the LAION training-data findings), and its transparency reporting is the main published evidence of enforcement. The address is real; its throughput is unknown.

Notes

safety@stability.ai is the route. Note there is no frontier safety framework or scaling policy — the AUP and the six Stable Safety principles are the whole published governance surface. For open-weight models, reporting misuse to the developer has limited effect once weights are distributed; that is a structural limit, not a channel failure.