According to Reuters, a helium shortage linked to the Middle East conflict has started to affect global tech supply chains. The report says helium is used in cooling, leak detection and precision chipmaking processes, so the squeeze is now reaching semiconductor production directly.
Reuters reports, South Korea has approved a 250 billion won investment in AI chip startup Rebellions. The funding is presented as the first direct investment under the state-backed ‘K-Nvidia’ initiative and is meant to support mass production of NPU chips and next-generation AI semiconductors.
According to Reuters, Shield AI is raising $2 billion at a $12.7 billion valuation to expand its AI-powered defence software. The article says demand has risen with wider battlefield use of autonomous tools, and Shield’s Hivemind software is already being tested on F-16s and the US Air Force’s Collaborative Combat Aircraft programme.
Regulation
- The Department for Science, Innovation and Technology has published ‘Deepfake detection technology’ on 26 March 2026. The paper describes deepfakes as AI-generated or AI-manipulated synthetic media, says the UK market is still at an early stage, and sets out barriers, demand conditions and growth scenarios for detection tools.
- The UK government has published the ‘Smart Data 2035 Strategy’ on 26 March 2026. The strategy says future smart data schemes should work with the wider UK data ecosystem, ‘including artificial intelligence (AI)’, which places AI use inside the government’s longer-term data access architecture rather than treating it as a separate policy track.
- The Scottish Government has issued Scotland’s AI Strategy 2026-2031, published on 20 March 2026. The strategy is framed as a five-year plan to use AI for ‘responsible and inclusive growth’, and its action plan includes an AI regulation layer alongside infrastructure, data and adoption.
Cases
- Via CourtListener, it can be seen that In Re OpenAI, Inc. Copyright Infringement Litigation had a same-day docket update concerning sealing. The judge granted the motions to seal, keeping confidential material on OpenAI’s data retention and output log practices out of the public record.
Academia
- SSRN has posted ‘The Surprising Virtues of Heterogeneity: Legal Pluralism and the Governance of Generative AI’. The paper argues that fragmented or plural legal approaches can be a strength in governing generative AI rather than simply a coordination failure.
- arXiv has posted ‘Runtime Governance for AI Agents: Policies on Paths’. The paper treats the execution path as the main object of governance and argues that prompt instructions and static access controls are not enough for path-dependent agent behaviour.
Events
- Goldsmiths, University of London lists ‘Statecraft, Sovereignty and Digital Government’ for 16 to 17 April 2026. The symposium explicitly includes ‘AI and digital government’ among its themes, placing public-sector AI within wider debates on sovereignty, digital transformation and state capacity.
Takeaway
The official material today points to a more operational phase of AI governance. The state is not only discussing principles but mapping concrete markets such as deepfake detection, fitting AI into broader data policy, and tracking how AI disputes are maturing into durable litigation.
Sources: Reuters, GOV.UK, Scottish Government, CourtListener, SSRN, arXiv, Goldsmiths