The EU AI Act's new dates for non-EU suppliers.
Regulation (EU) 2026/1744 moved Annex III high-risk duties to 2 December 2027 and Annex I to 2 August 2028. What applies now to AI suppliers outside the EU.
resources
Insights.
Field notes on shipping AI in regulated work.
All posts
Offshore AI suppliers and data in your region.
Remote access from outside the EU or UK can be a data transfer, even if data is only shown on screen. What the EDPB and the ICO say, and what to ask suppliers.
Australia Ask an AI vendor for evidence you can check.
Questions to ask before you buy or renew an AI product, drawn from APRA, the DTA, NSW buying guidance and the OAIC, and what a good vendor answer looks like.
Australia Find the automated decisions in your software.
From 10 December 2026, privacy policies must name decisions a computer program makes or substantially assists. How to find yours, licensed software included.
Australia Copilot is only as safe as your settings.
Where Microsoft 365 Copilot stores and processes Australian prompts, what goes to Bing, and what lawyers, tax practitioners and clinicians should check first.
Australia Measure Copilot before the licences go out.
Time 3–5 recurring tasks and their error rates before Copilot licences go out, then compare on a set date. What the government trial and Microsoft reports miss.
Australia NSW agencies need AI evidence from suppliers.
What circular DCS-2026-02 and the NSW AI Operational Policy require of agencies, and the testing, change and record evidence they'll need from AI suppliers.
Define the scope of an AI automation project.
Understand the work behind an AI automation project: discovery, access, document preparation, approval controls, testing, handover and ongoing responsibilities.
Choose your first AI workflow.
Choose a useful first AI automation project: one recurring task, accessible inputs, a reviewer and a measurable baseline, with a synthetic control walkthrough.
Australia Your practice software already runs AI.
What JAX, MYOB AI BAS, Dext and Karbon AI do, whether each vendor says where its AI runs, and what the TPB asks before client information goes into any AI tool.
Australia Test a Copilot agent before it leaves pilot.
What the DTA policy, NSW's AI circular and Queensland's FAIRA ask of a Copilot agent leaving pilot, and what an independent test adds to Copilot Studio's own.
Australia Courts want a person to verify AI output.
NSW SC Gen 23, Victoria's SC Gen 25, the County Court's note and the Federal Court's GPN-AI put verification on a person. What to test in a legal AI tool first.
Australia Check where property AI sends tenant data.
What PropertyMe's AiMe features and Rex AI do, what each vendor publishes about where its AI runs, and the renter-data rules to check before you switch them on.
Australia Test school AI before staff use it.
The Australian Framework for Generative AI in Schools expects AI tools to be tested before use and monitored after. What that asks of a staff policy assistant.
Australia Re-test SOP answers before each model update.
Safe Work Australia's AI guidance says vendor-pushed updates may change WHS risks. How to test an SOP assistant in each language, before and after each update.
Australia Measure what an AI scribe leaves out.
The TGA, ACSQHC and Ahpra ask for AI scribes to be checked before use and reviewed after updates. How to test one on scripted consultations in your specialties.
Australia A fictional utility tests its SOP assistant.
An illustrative scenario: a made-up water utility tests a procedures assistant on its crews' questions and maps where schematics would go under the CIRMP Rules.
Australia Re-test vendor AI whenever the model may change.
APRA found few of the large entities it reviewed validating AI continuously. How a test set from your own cases shows your second line what a model change did.
Australia AI hosting in Australia: check the data path.
Compare self-hosted models, Amazon Bedrock and Azure OpenAI. Check routing, storage, access and model availability before choosing an Australian deployment.
Australia Shipping a RAG system that stands up to APRA scrutiny
Grounding, inline citations, retrieval evals, and a refusal path — the engineering that makes a retrieval system answer to an auditor, not just a demo.
Evals before vibes: release gates for LLM systems
How we stop 'looked good in the demo' from reaching production — the offline eval suite, the regression gate, and the tracing we wire in from day one.