Overview
Severity: MEDIUM | Affected: EU Commission | Category: policy
The European Commission has officially ratified the AI Safety & Accountability Act (ASAA), introducing stringent new regulations for developers and operators of high-risk AI systems. The landmark legislation, set to be enforced from Q4 2026, mandates two key provisions. First, it requires companies to maintain immutable, auditable logs for the entire lifecycle of their high-risk AI models, covering training data provenance, architectural changes, and inference decisions. This is intended to provide transparency and aid in forensic analysis after a safety incident. Second, the ASAA legislates a mandatory 'red-button' protocol, requiring a secure, verifiable mechanism for the immediate shutdown of autonomous systems that operate outside of pre-defined safety boundaries. Non-compliance will result in fines of up to 5% of a company's global annual turnover. The act is seen as a significant step in establishing global AI governance standards, forcing companies worldwide to re-architect their systems for compliance if they operate within the EU market.