AI Hotlist: 3 to know this week – Thursday 25 June 2026
This week’s stories focus on China’s open‑weight push, labs moving into custom chips, and Europe trying to build a “regulated and open” AI path.
1️⃣ China’s GLM‑5.2 crashes the coding party
👀 What happened: Chinese lab Zhipu AI has released GLM‑5.2, an open‑weight large language model that’s rapidly becoming a favourite for code and long‑context work. It offers a huge context window and can be downloaded and self‑hosted, putting it in direct conversation with top‑tier US models for software engineering tasks rather than just chat or simple Q&A.
❓Why it matters: For developers, startups and enterprises, GLM‑5.2 shows that powerful, open‑weight models from China are now a credible option for serious coding and agentic workflows – but they come with new questions around governance, compliance and geopolitical exposure when you embed them deep into your stack.
🔗
hubs.li/Q04mCPMX0
2️⃣ OpenAI joins the AI‑chip arms race
👀 What happened: OpenAI has unveiled “Jalapeño,” its first custom AI chip, built to run ChatGPT‑class systems faster and more cheaply in data centres. Rather than relying solely on off‑the‑shelf GPUs, the lab is moving down the stack into silicon, co‑designing hardware that’s tuned specifically for its own models and inference workloads.
❓Why it matters: For hyperscalers, chipmakers and enterprise buyers, this is a clear step toward vertically integrated AI stacks where the same players control models, infrastructure and now custom silicon. It tightens competition with Nvidia, reshapes cost structures for large‑scale deployments, and signals that leading labs see proprietary hardware as a strategic moat, not just a performance tweak.
🔗
hubs.li/Q04mCM2H0
3️⃣ Europe doubles down on a “regulated and open” AI path
👀 What happened: As the EU’s AI Act moves from text to implementation, member states are starting to stand up national enforcement regimes and support home‑grown models. Alongside new domestic AI bills, Brussels is backing efforts to build an open‑source, frontier‑grade model that serves all official EU languages and can be scrutinised, audited and adapted within the European regulatory framework.
❓ Why it matters: For policymakers, enterprises and AI builders working with European customers, AI is now firmly a compliance and sovereignty issue as much as a technology one. The emerging offer is a distinct “third way”: tightly regulated, rights‑driven AI built on more open foundations – which could become both a constraint and an advantage for companies that are ready to align with it early.
🔗
hubs.li/Q04mCDxS0
Was this useful? If so, follow along and we’ll keep tracking the stories that matter in AI.
Want to share a story you think the WSAI community should be paying attention to?
Drop it in the comments with your take. ⬇️
#WorldSummitAI #AIHotlist #AIGovernance
