Minister Solly Malatsi is committed to having a new draft AI policy in front of the Cabinet by the end of the financial year (March 2027), with the goal of one free from avoidable stains, and is running the rewrite through an expert advisory panel led by Prof Benjamin Rosman of the Wits Mind Institute to reconstruct a policy befitting the unique South African reality, after the previous draft was withdrawn for containing AI-hallucinated citations.
“Mapping U.S. Federal AI Governance Against Sector Vulnerability” reports that an assessment of 684 federal AI governance documents finds substantial variation in coverage of 14 sectors and 24 AI risks, with AI risks related to robustness, system security, and governance receiving more attention than socioeconomic, environmental, and emerging risks including multi-agent risks, and with public administration, national security, information, and scientific services receiving higher coverage than finance and healthcare, which experts rate as highly vulnerable to AI risks.
Al Issaei pointed to a diabetic retinopathy screening project involving more than 35 healthcare institutions, where AI is used to analyse retinal images and support the identification of possible problems, and Chapu said that imperfect data does not necessarily prevent governments from deploying AI for advisory applications, where a human remains responsible for the decision.
Biometric Update reports that Americans reported a record $15.9 billion in scam losses to the Federal Trade Commission in 2025, up 25 per cent from the previous year, while an Associated Press and FRONTLINE investigation found that AI increasingly enables criminal organisations to operate at greater scale and sophistication, part of the case for a cryptographically verifiable credential whose trust rests on mathematical proof of who issued it and whether its contents have since been altered.
The authors of “Governance-as-Code: Translating EU AI Act Technical Requirements into Executable Compliance Pipelines for Generative AI Systems” report that they deliver a framework of 43 machine-checkable acceptance criteria across six compliance modules that run in a CI/CD pipeline and emit Article-indexed audit evidence, and the paper publishes the actual policy code alongside the description.
Today’s lead is ITU’s GENIE.AI is an open-source AI stack now piloting in four countries.