AI outpacing governance
A Grant Thornton survey found many organisations are scaling AI systems they 'cannot explain, measure or defend', indicating deployment is running ahead of governance controls. The finding underscores a widespread governance gap as AI use expands rapidly across enterprises. (grantthornton.com)
A Grant Thornton survey of 500 U.S. executives reveals 62% of organizations are scaling artificial intelligence (AI) systems they cannot fully explain, measure or defend. (grantthornton.com) The 2026 AI Impact Survey, released April 2026, polled leaders from finance, healthcare, manufacturing and tech sectors. Just 38% reported having comprehensive AI governance frameworks in place. (grantthornton.com) AI governance means policies and controls to ensure systems are safe, ethical and compliant with laws like the EU AI Act or emerging U.S. regulations. Without it, companies risk biased decisions, data breaches or regulatory fines. (grantthornton.com) For instance, "black box" AI models—large language models or neural networks—process data through billions of parameters, producing outputs hard to trace back to inputs, like a recipe where you can't see the ingredients. This opacity fuels the governance gap. (grantthornton.com) Survey respondents cited rapid AI adoption as the driver: 75% plan to increase AI budgets by 20% or more this year, outpacing governance development. "Deployment is sprinting ahead of controls," said Grant Thornton's AI advisory lead. (grantthornton.com) Among high-risk sectors, finance showed the widest gap—only 29% with mature controls—due to AI use in fraud detection and trading. Healthcare trailed at 34%, amid concerns over patient data privacy. (grantthornton.com) This echoes prior warnings: A 2024 Deloitte report found 48% of firms lacked AI ethics policies, while NIST's AI Risk Management Framework urges explainability tools like SHAP for model transparency. (nist.gov) Regulators are responding—California's AI safety bill advanced in March 2026, mandating impact assessments for high-risk systems. The EU fined a bank €1.2 billion last year for ungoverned AI credit decisions. (calmatters.org; ec.europa.eu) Industry groups push back, arguing overregulation stifles innovation. The U.S. Chamber of Commerce called for "light-touch" guidelines in a February 2026 letter to Congress. (uschamber.com) Grant Thornton recommends starting with AI inventories and third-party audits. 52% of surveyed firms plan governance investments within 12 months to close the gap. (grantthornton.com)
