Auditing AI once isn't enough. That's the premise behind two new technical committees announced today by Codema, the Swiss standards body, which are set to work on a framework that embeds compliance checks into AI systems as they run — not just at deployment.
TC23 and TC24, the two new committees, tackle complementary problems. TC23 will focus on assessing AI system performance: reliability, traceability, and anomaly monitoring. TC24 covers the human side of that same picture — how AI integrates into business processes, what human oversight looks like in practice, and how evidence of compliance is collected and verified.
The 'Audit as a Layer' approach Codema describes is a departure from the dominant model in AI governance, which remains point-in-time: a system is assessed, certified, and then assumed to be operating correctly until the next audit cycle. As AI models update, retrain, and interact with changing data distributions, that snapshot approach leaves significant gaps.
Codema draws on its origins in pharmaceutical-sector assessment, a field that requires continuous process monitoring rather than periodic certification. The methodology has since been extended to AI, DLT technologies, digital infrastructure, sustainability and life sciences.
The McKinsey AI Trust Maturity Survey 2026 provides the market context Codema cites: only 30% of organisations have reached advanced AI governance maturity. The gap between AI deployment and credible oversight infrastructure is the opening Codema is targeting.
The committees will work within what Codema calls a 'stakeholderisation' model, drawing in businesses, auditors, certification bodies, academics, and technology experts to develop shared criteria and audit protocols. The output is intended to be adoptable across industries, not specific to any sector.
To stay across the latest in cloud, AI, and enterprise tech analysis from Compare the Cloud, subscribe to our weekly newsletter at comparethecloud.net/newsletter
