RELAY
Subscribe

Relay Weekly · No.2

Sovereign AI in APAC, chatbot disclosure, and Gemini's enterprise gap

What sovereign AI means in practice, what to know about IMDA's transparency guidelines for chatbots, and why Gemini's consumer lead is not enterprise readiness.

Shane Neubronner · 14 August 2026

Subscribe to this newsletter or send story tips to hello@relay-asia.com.

This Week's Analysis

What sovereign AI actually means in APAC

Welcome to Relay Weekly No.2.

This month Indosat, Nokia and Nvidia launched Zankore — a full-stack AI infrastructure platform pitched at Southeast Asia, targeting up to one gigawatt of AI factory capacity, with an initial 200 megawatts targeted for the first half of 2027. Justin Hotard, President and CEO of Nokia, put the phrase on the record: countries and enterprises are increasingly focused on building sovereign AI capabilities that are secure, scalable and trusted — with compute, connectivity and control orchestrated across AI factories, data centres and the edge.

So what does sovereign AI actually mean? This week's Analysis breaks this down and looks at how sovereign AI in APAC is not about developing every aspect of AI infrastructure domestically, and at the different approaches Taiwan, Vietnam, Malaysia and Singapore take to language, data, compute, regulation and partnerships. Most markets still buy chips, models and cloud services from abroad, yet there are important choices on what they need to control, such as where data is held, who provides the compute, and which rules apply.

Read on Relay Asia →

Short Reads

  1. Why Singapore's GenAI chatbot transparency rules may offer an APAC disclosure reference

    Details about what a chatbot does, its limits, safety measures and privacy handling often sit across product pages, terms and notices. This Short Read looks at Singapore's voluntary chatbot information card, which brings that information together where a person can find it at first use, and why the card only helps if it stays current when the product changes. It also compares how Australia, Japan and South Korea approach disclosure at different layers.

    Read on Relay Asia →
  2. Singapore's Gemini per-capita lead — and what it may mean for enterprise AI

    Google's 2026 Gemini Southeast Asia report ranks Singapore first globally for Gemini adoption per capita. This Short Read looks at what that consumer-app data actually measures, and why it is not the same as enterprise readiness. It sets the ranking against Singapore Ministry of Manpower figures on firm adoption, and asks what an enterprise view would need to show: redesigned work, governed staff use, and AI in core processes.

    Read on Relay Asia →

Suggested Reads

  • The Non-Technical Technical Dictionary — Good things are worth sharing. If you're struggling with new tech terms, or need to brush up on terminology, here's a site with plain-English explainers for the language that keeps showing up in AI and product conversations: application programming interfaces (APIs), tokens, context windows, Model Context Protocol (MCP), agents. Worthy of a bookmark.
  • Why full AI self-sufficiency would hold Asia Pacific back — Read this after our Analysis as Oxford Economics models what happens when governments push for a fully domestic AI stack instead of a sovereign mix of local control and global technology — adoption can slip by three to five years, and the direct costs run to tens of billions in the larger economies.
  • South Korea plans $710m AI sovereign wealth fund — Seoul is putting state money behind AI, data centres and robotics through Korea Investment Corp. Another perspective on sovereign AI as a capital strategy — investing in the sector rather than trying to build every layer of the stack at home.
  • NUS makes AI courses compulsory and rolls out ChatGPT Edu campus-wide — The Straits Times reported on 11 August 2026 that the National University of Singapore (NUS) is making applied generative AI compulsory for undergraduates, and will give the whole campus ChatGPT Edu from 31 August to build literacy.
  • Privacy concerns over Western Australia's live AI police face-scanning trial — Western Australia Police have been running a live facial-recognition trial since June. ABC reported on 11 August 2026 that cameras scanned more than 130,000 people in Perth and Fremantle in the first week, matching faces against a watchlist of about 4,000. Privacy advocates are asking how that list was compiled, and where the cameras sit.
  • Thailand accelerates AI government with law and data reform — On 6 August 2026, Deputy Prime Minister Pakorn Nilprapunt argued that Thailand's public-sector AI push depends on legal reform, machine-readable data across agencies, and civil-service literacy. He also shared how Thailand is building the digital government ecosystem.

For a clearer filter on Asia's technology, business and policy stories, covering what changed, what it means and what to watch, stay on this mailing list. Forward the edition to one colleague who is tired of undifferentiated AI coverage. And if you see something we should dig into, reply or write to hello@relay-asia.com.

— Shane

← All Weekly