IBM Granite

Stav’s assessment

Editorial assessment, not legal advice. Stav's risk ratings, scores, and verdicts are our own analysis of publicly available information and may be incomplete or out of date. Verify independently before making compliance or procurement decisions.

• transparency

  IBM publicly discloses Granite pretraining datasets and sampling methodology, explicitly rejecting industry trends toward concealment of training data.

• security

  As of 29 April 2026, IBM cryptographically signs released Granite language, vision, speech, embedding, and guardian models, adding a verifiable provenance layer.

• compliance

  IBM has signed the EU GPAI Code of Practice, a Commission-endorsed voluntary tool providing a presumption of conformity with EU AI Act GPAI obligations.

• community

  Granite models are distributed as Apache 2.0 open weights via HuggingFace, Ollama, NVIDIA NIM, and Google Vertex AI, with an active HuggingFace org showing frequent recent updates.

• stability

  IBM is a long-established, financially stable public corporation with a continuous, rapid Granite release cadence (3.0 through 4.1) and no signs of contraction or leadership disruption to the programme.

Published safeguards & certifications