News
AI regulation and accountability, as it happens
Policy updates on AI law in South Africa and beyond, each with its primary source, and analysis from AIC on what they mean for the people who answer for automated decisions.
Analysis, Best Practices
What SIL can and cannot tell you about AI
Functional safety engineering has spent decades building a rigorous language for how much a safety function can be trusted. It is tempting to extend that language to AI components. Some of it transfers well. The part everyone wants most does not transfer at all.
AIC, 2 September 2026
Analysis, Certification
Why our register is empty
AIC's public register currently lists no certified organisations, because none are certified. We could have populated it with examples, pilots or logos. Here is why we did not, and what the design of an empty register commits us to later.
AIC, 2 September 2026
Analysis, Policy Updates
Six citations that did not exist
South Africa withdrew its draft National AI Policy because six of its academic references were invented. The instructive part is not that a machine produced them. It is that the document travelled from drafting through internal review, Cabinet approval and publication in the Government Gazette without anyone checking.
AIC, 2 September 2026
Policy update, Guidance
Synthetic content marking: grace period closes 2 December 2026
The EU AI Act's requirement to mark AI-generated audio, image, video and text in a machine-readable format has applied since 2 August 2026, but systems already on the market were given until 2 December 2026. That is the nearest live deadline in the Act and it is three months away. Marking is a build task, not a policy one, which is what makes it easy to leave too late.
2 September 2026 for European Union
Policy update, Regulatory
Article 50 transparency obligations now apply
From 2 August 2026 the EU AI Act's transparency duties apply: people must be told when they are interacting with an AI system, synthetic audio, image, video and text must be machine-readable as artificially generated, those exposed to emotion-recognition or biometric categorisation must be informed, and deepfakes and AI-generated text on matters of public interest must be labelled. Penalties reach EUR 15 million or 3% of global turnover. One grace period remains open.
2 August 2026 for European Union
Policy update, Regulatory
Digital Omnibus in force: high-risk AI obligations deferred to December 2027
The EU's Digital Omnibus on AI was published in the Official Journal on 24 July 2026 and entered into force on 27 July. Obligations for stand-alone high-risk systems (Annex III) move from 2 August 2026 to 2 December 2027, and for AI embedded in already-regulated products (Annex I) to 2 August 2028. The deferral is now law rather than proposal, which matters: many compliance plans were still being written against the original August 2026 date.
27 July 2026 for European Union
Policy update, Regulatory
South Africa's draft National AI Policy withdrawn in its entirety
The Draft National AI Policy, gazetted on 10 April 2026, was withdrawn in full by government notice on 12 June 2026 after at least six of its 67 academic references were found not to exist. Journal editors confirmed the attributed articles had never been published. Two officials are on precautionary suspension and an internal investigation continues. Revised documents are targeted for Cabinet in November 2026, with public comment in January 2027.
12 June 2026 for South Africa
Get new updates by email
One email when we publish a policy update or a piece of analysis. No marketing sequence, and you can leave with one click. Prefer a feed reader? Use the RSS feed.