Skip to main content

AI Governance & Safety

Model registry routing, circuit breakers, retrieval controls, confidence gates, and human acceptance checkpoints.

In Plain Language

AI safety in ESG reporting requires governance, not guesswork. The model registry routes tasks to the best-suited LLM. Circuit breakers prevent cascading failures. Retrieval accuracy gating ensures AI claims are grounded in source paragraphs, not training data.

Technical Requirements

  • Model registry & routing governance
  • Circuit breaker & failover architecture
  • Retrieval accuracy gating thresholds
  • Human acceptance checkpoint workflow

Sources

  1. IFRS S1 §54-59Sources of guidance (Part_A_Standards/IFRS_S1_General_Requirements_for_Disclosure_of_Sustainability-related_Financial_.pdf)
  2. IFRS S2 §27-36Climate metrics and targets (Part_A_Standards/IFRS_S2_Climate-related_Disclosures.pdf)
  3. IFRS S1 §21-24Connected information (Part_A_Standards/IFRS_S1_General_Requirements_for_Disclosure_of_Sustainability-related_Financial_.pdf)