AI Scaling Outpaces Governance and Security, New Reports Warn

In Zambia, health workers using an AI-guided mobile app detected cholera signs months before confirmed cases appeared, according to a TechRadar report. The app, developed by Australian team ThinkMD, used clinical prompts to turn routine primary-care consultations into syndromic surveillance. In Abu Dhabi, a similar approach alerted officials that the influenza season would arrive early, allowing them to launch vaccination campaigns ahead of the usual schedule, the report said.

AI's impact on software is equally pronounced. Veracode's 2026 State of Software Security report, covered by TechRadar, found that 82% of organizations carry security debt, and 60% have critical security debt. Third-party code represents 66% of the most dangerous, long-lived vulnerabilities. The report argued that AI accelerates code creation and deployment, widening the gap between software creation and remediation.

The governance gap extends beyond security. IBM Consulting managing partner for UK and Ireland told TechRadar that AI adoption is evolving from isolated experiments to enterprise-wide deployment, making governance a central concern. The UK government's AI Opportunities Action Plan, cited in the same piece, estimates that AI could boost UK productivity by 1.5 percentage points annually, generating £47 billion in economic gains each year. It also highlights £14 billion in private-sector AI investment commitments and more than 13,000 planned jobs. A poll found 72% of the British public would be more comfortable with AI if laws and regulation were in place.

For nonprofits, the speed of adoption is less pressing than the need for careful implementation. Enterprise software firm Unit4 advised the sector to pursue incremental gains rather than radical transformation, warning that a "Big Bang" approach increases the risk of failure. A Unit4 study found that 61% of US nonprofit finance professionals still use generic spreadsheets for core financial management, complicating data governance. Without proper data setup, AI tools may misinterpret terms or generate inaccurate results, the company said.

AI's limitations also persist in testing. MIT Technology Review reported that even top-tier models fail at spatial reasoning and visual puzzles, and subtle changes to classic riddles can trip them up. While models improved from solving 18% of New York Times Connections puzzles in late 2024 to near-perfect performance in early 2025, they still rely on memorization rather than robust reasoning in some cases.

Some productivity gains attributed to AI may have simpler explanations. Bloomberg reported that a 2015 plan for Andy Burnham's Labour leadership bid warned of a "productivity problem" in Britain, predating the current AI push. That suggests the country's recent efficiency improvements are not necessarily driven by AI.