AI Governance and Regulation Face Global Scrutiny Across Sectors

A new study by Cornell and Carnegie‑Mellon researchers warns that weak AI safety regulations can backfire, arguing that effective safeguards must target model developers such as OpenAI, Google and Anthropic rather than only downstream users. Separate research from French and Italian scholars found that AI suggestions dramatically reduce people’s willingness to admit they don’t know an answer, cut correct‑answer rates from 27 % to 9 % while inflating confidence, raising concerns about critical‑thinking erosion. Enterprise reports show that most organisations lack visibility into the AI tools they run; only 13 % have full oversight, and 71 % have experienced audit or compliance failures linked to AI‑driven governance tools. Industry analysts highlight the need for layered AI security, citing gaps in real‑time governance, developer‑focused risk controls and the upcoming Model Context Protocol update that will streamline AI‑app integration. Healthcare leaders stress the urgency of robust guardrails as AI expands into clinical decision‑support, while education and business groups launch training programmes to build AI skills and mitigate adoption risks. Companies such as Box are adding agent‑level approvals and audit trails, and broader market surveys reveal that while AI is widely deployed, many firms struggle to embed it into workflows and realise revenue growth.