The Agent Watch
Briefing · Articles · Tools · About EN FR DE ES 中文 IT PT SV FI DA

Daily Briefing

24 August 2026 · 5 stories

🔥 Top story

1

Alibaba raises $10 billion for AI: China opens its chequebook

Imagine a company selling a slice of itself to pocket $10 billion in one day, and promising to spend it all on artificial intelligence. That's what Alibaba did on Sunday 23 August in Hong Kong. It's the largest share placement in the history of the Hong Kong Stock Exchange, and the third biggest in the world in 2026. To put it in perspective: that's more money than it takes to buy 100 Airbus A320s. Alibaba will use it to make its own chips, build data centres and train its AI models. The stock still dropped 10% on Monday morning at the open — the market is digesting the dilution, meaning each share is now worth a little less because there are more of them. For the rest of us: China is taking the lead, and the others will have no choice but to follow. Source: https://www.reuters.com/business/alibaba-10-2b-hk-share-placement-ai-2026-08-23

2

Chinese model GLM-5.3 will be open-source at the end of August — and it can code

In China, lab Z.ai has announced it will release, around 28 August, the weights of its GLM-5.3 model — the internal engine that powers the AI. "Open weights" means anyone can download it for free and run it at home, without paying Z.ai a subscription. It's a bit like Renault opening up the plans of its electric engine to every mechanic in the world. The model is particularly good at code: according to Z.ai, it scores 50% better than the previous version on programming exercises, and it tops the open-source leaderboards on the reference benchmarks. One spicy detail: Z.ai also published the list of 2,436 security vulnerabilities it found in open-source software, including the Linux kernel — proof that they really put their model through the wringer before releasing it into the wild. Source: https://z.ai/blog/glm-5-3-weights-release-window-2026-08-14

Source : z.ai →
3

OpenAI now makes $40 billion a year — its revenue has doubled in 7 months

When we say OpenAI is making $40 billion a year, we mean that if you project what the company earned in recent months over a full year, you get to $40 billion. To put that in perspective: it's more than the annual revenue of most large European banks. And seven months ago, that figure was $20 billion — so it has doubled. The growth drivers: paid ChatGPT subscriptions, the new business of integrated advertising, and above all the tools for enterprises. More than 40% of revenue now comes from large companies using OpenAI's APIs to automate their processes. For the rest of us: AI is no longer a gadget for techies, it has become a real business at industrial scale, and OpenAI is clearly preparing to go public in the coming months. Source: https://www.bloomberg.com/news/articles/2026-08-13/openai-annualized-revenue-40-billion

4

European AI rules are now in force: what changes for you and your company

On 2 August 2026, the European Union lit the fuse: the strict rules on "high-risk" AI came into force. This concerns all AI systems used in hiring, healthcare, banking, justice and education. In practice, your company must now document how its AI works, keep logs for six months, prove that a human supervises the decisions, and report any serious incident within 15 days. If it doesn't, the fine can go up to 15 million euros or 3% of global revenue — whichever is higher. And it's not just in Europe: if your AI is used by someone living in the EU, you're covered, even from Norway or the United States. For 78% of organisations, these rules are a surprise — the compliance window is open, and experts expect the first fines by the end of the year. Source: https://artificialintelligenceact.eu/article/50/

5

Three cloud giants launch in 24 hours the same recipe to run AI agents in production

Imagine you want an AI agent to work for you 24/7, on its own, without you watching. Until now, this was hard to set up. But in 24 hours, three cloud giants — Snowflake, Amazon and Google — rolled out almost the same solution: a protected space (a "sandbox") where the agent can run on a loop, triggered by a schedule or an event, and where you can inspect every decision afterwards. It's a bit like if Netflix, Disney+ and Amazon Prime Video all released, on the same day, the same new streaming format — a sign that it has become the standard. For an SME, this means launching an autonomous agent becomes a project of a few days, not several months. One caveat: Snowflake leaves the execution rights open to all by default — you need to remember to restrict them. Source: https://www.snowflake.com/blog/coco-automations-public-preview-2026-08-21

📡 To watch

The rest of the Alibaba placement: where does the money really go?

The placement should be finalised this week. The split between chips, data centres and models will be a strong signal for the sector — watch for announcements from Morgan Stanley, HSBC and UBS.

GLM-5.3 open-source: it's coming very soon

Z.ai has promised the open weights release around 28 August. Watch the lab's Hugging Face page "zai-org" and the z.ai blog in the 26-30 August window. If the licence really is MIT, it's a real revolution for European users.

OpenAI and Anthropic are getting ready to go public

Both labs have filed confidential IPO paperwork. Bloomberg and The Information are watching for leaks of the S-1 document — the first leak is expected within 30 days.

First European AI fines by the end of 2026?

22 days after the deadline, European regulators have not yet struck. But 78% of organisations are still not in compliance — the window to get into compliance without pain is closing. Worth watching to calibrate your strategy.

📊 Trend

The week ahead is dominated by money: $10 billion raised by Alibaba on Sunday, $40 billion in annualised revenue for OpenAI, and a new 'agent in production' standard adopted by three cloud giants in 24 hours. For the first time, European AI rules are no longer text to come: they apply, and the first fines are coming by the end of the year. Three forces are converging — massive capital, a standardising technical stack, and a tightening legal framework — and they will decide who leads the race in the next twelve months.