A provider of a general-purpose AI model must determine, and keep current, whether the model is a model with systemic risk. It is, where it has high impact capabilities evaluated on the basis of appropriate technical tools and methodologies including indicators and benchmarks, or where the Commission so decides on the criteria in Annex XIII. Cumulative computation used for training greater than 10^25 floating point operations creates a presumption of high impact capabilities. The thresholds, benchmarks and indicators may be amended by delegated act, so the assessment must be made and maintained against the criteria in force.
This control maps to 2 controls across 1 other frameworks. If you already hold one of them, the evidence you collected for it is the starting point here rather than new work.
Every mapping shown was judged rather than inferred from wording similarity, and the ones that failed review are published too. See the coverage reports and what was rejected.
EU AI Act EUAI-Art.51 is one control. If you already hold one of the frameworks below, a reviewed crosswalk already says how much of EU AI Act your existing evidence covers. Hold ISO/IEC 42001:2023 and 17 of 43 EU AI Act controls already carry evidence.
Each report names every control your existing framework evidences, every one it does not, the reasoning behind each claim, and the claims that were argued against and rejected. 0 were rejected on the ISO/IEC 42001:2023 pair alone.
The graph holds this control, the 2 it maps to, and the evidence behind each claim, over MCP and REST.