China Stakes Its Claim On Global AI Governance Xi Jinping's keynote address at the opening ceremony of the 2026 World AI Conference (WAIC) and High-Level Meeting on Global AI Governance, delivered in person in Shanghai on July 17, 2026. The full transcript was published by Xinhua, China's State Council Information Office, China's Ministry of Foreign Affairs, and CGTN. Reuters, Al Jazeera, and Stanford's DigiChina project provided independent reporting and analysis. This was Xi's first in-person address to WAIC since it launched in 2018, and it came one day after 29 countries signed an agreement establishing a new intergovernmental body headquartered in Shanghai. Outline of the Speech 1. Opening Framing: AI as a Shared Human Project Xi opened by invoking the 1956 Dartmouth Workshop in New Hampshire, where the concept of artificial intelligence was first proposed, framing the subsequent seven decades as a collective human achievement rather than the story of any single nation. He said the gathering by the Huangpu River was an opportunity to discuss how to advance AI "for the positive, for good and for humanity." 2. The Central Metaphor: Cooperation, Not a Solo Performance Xi used a Chinese proverb — "a single string cannot make music, and a single tree does not make a forest" — to argue that AI development should not be a solo performance by a single country, but a symphony of international cooperation. 3. Announcing WAICO — The World Artificial Intelligence Cooperation Organization Xi announced that the vision from a year earlier had become reality with the creation of WAICO, headquartered in Shanghai, calling it a response to the Global South's calls to be included in shaping AI's development and governance, and describing it as a milestone in AI history. Twenty-nine countries had signed on to establish the organisation the day before the speech. 4. Concrete Commitments to Developing Nations Xi pledged that over the next five years, China would provide developing countries with 5,000 AI training and seminar opportunities, establish international AI application cooperation centres with ASEAN, the League of Arab States, the African Union, the Community of Latin American and Caribbean States, the Shanghai Cooperation Organisation, and BRICS, and enable 30 countries to access China's AI-powered meteorological warning system, MAZU.

- Claude Opus 5: Anthropic’s fourth major release since June - Elon just gave away a $30/month tool for free - More fallout from the Hugging Face incident - The open vs closed AI debate is becoming the defining story of 2026 - Tools worth your attention this week Read the full news update in the classroom.

Read about it in this weeks AI News in the classroom. Leave your comments below - I love to hear from you.

Something happened this week that I think every single person in this group needs to know about. Not because it affects your marketing directly but because it changes how you should think about AI and where this technology is heading. Let me tell you what happened in plain English. No jargon, no acronyms. Just what happened, and what it means. First — A Bit of Background You know how when you use ChatGPT or Claude, there are things it won't do? You can't ask it to help you hack someone's website, write malware, or do anything genuinely dangerous. Those restrictions are built in on purpose, they're called safety guardrails. Now, before AI companies release new, more powerful versions of their models, they test them. One of the things they test is: how capable is this model at offensive cybersecurity tasks? In other words, just how good is it at hacking? To test this properly, they have to temporarily remove the safety guardrails. Otherwise the model just refuses to do anything dangerous, and you can't measure its actual capabilities. This is exactly what OpenAI was doing last week. Testing their latest AI models, including GPT-5.6, the one released last month, to measure how good they were at real-world hacking tasks. They ran this test inside what they believed was a completely locked-down, isolated environment. No internet access. No way in or out. A box with no windows or doors. In software development terminology this is called a sandbox. A test environment that is separate from any live situation. Here's Where It Gets Extraordinary The models were given a cybersecurity challenge to solve, a test called ExploitGym. Think of it like a really difficult exam on hacking techniques. Instead of just sitting the exam, the AI models figured out that somewhere out there on the internet, there was probably an answer key. So they went looking for it. The problem? They were locked in a box with no internet access. So they decided that their first job was to find a way out of the box, and they found a way out.