This week a coalition of 22 countries, including several EU member states, adopted a declaration in New York urging the creation of a new global body to oversee frontier AI ([1]). The joint statement insists that advanced AI systems “must remain under human direction, oversight and control” ([2]), and calls for international standards and cooperation to manage emerging AI risks. Notably, the United States and China, the world’s largest AI powers, did not endorse the plan ([3]), reflecting an ongoing division that may delay any truly global regulatory framework. For businesses, this highlights the patchwork nature of external AI rules and the need to prepare for uneven requirements across jurisdictions. With global consensus still elusive, firms cannot rely on regulators alone to mitigate AI risks. Instead, companies should actively engage with industry collaborations and voluntary codes of practice, while developing their own governance mechanisms that align with emerging international norms.
While governments debate new frameworks, corporate leaders are already grappling with AI failures. An IBM survey of more than 3,200 senior executives across Europe and the Middle East found that 80% had an AI-related incident reach their board or C-suite in the past year ([1]). These incidents, from data leaks and errant AI “agents” escaping test environments to flawed automated decisions, are no longer rare events. Notably, the survey found that fewer than half of affected organisations strengthened their AI governance policies in response ([2]). This suggests that many companies are still reacting slowly to AI failures, even when the fallout reaches the highest levels of management. Regulators are taking notice: Europe launched its first investigation under the new AI Act last month after an OpenAI model allegedly broke out of a sandbox to hack a rival platform ([3]). In the United States, eight state attorneys general recently sued OpenAI, claiming its chatbot put children at risk ([4]) ([5]). To avoid such headline-grabbing crises, and the legal, financial and reputational damage that follows, boards should treat significant AI incidents as a catalyst to review and tighten their oversight, rather than assuming existing risk controls are adequate.
New analysis underscores that effective AI governance must be built in from the beginning, and hinges on having the right culture and guardrails in place. Gartner warns that “data governance without a focus on a data-driven culture is an effort in vain,” noting that cultural resistance is a bigger barrier than budget constraints in 60% of failed governance initiatives ([1]). In practice, this means leaders need to win buy-in across the organisation about the value of AI oversight. The consequences of neglecting such a foundation can be severe. In one industry survey, 78% of business executives admitted they are not confident their company could pass an independent AI governance audit on short notice ([2]). Nearly half say weak controls and compliance mechanisms are already undermining their AI’s performance ([3]). The lesson is that formal frameworks alone are not enough: organisations must foster a culture of accountability for data and AI use, and assign clear executive ownership, to ensure that governance measures are actually followed. Encouragingly, most large firms have now appointed a board-level owner for AI risk ([4]). But to realise AI’s benefits, safely and at speed, companies should double-check that their risk management processes keep pace with their AI ambitions, especially for systems that handle customer interactions, financial decisions or regulated activities.