24-Hour AI News: Qwen3-Max Released, Stargate Expansion, EU Action Against AI Fraud

In the past 24 hours (September 24th), Alibaba Cloud released the trillion-scale model Qwen3-Max and the new Wan 2.5-Preview. Supporting policy initiatives included the "AI+" international cooperation initiative and the release of a consensus on the ethics of AI services for minors. Overseas, OpenAI, Oracle, and SoftBank announced the addition of five new "Stargate" data centers. The EU inquired with several industry leaders about AI fraud management under the DSA. The UN high-level meeting placed AI governance at the core of its agenda. Meanwhile, Meta expanded Llama's use to allies in multiple countries. Overall, the amount of information flowing domestically and internationally remained balanced.

1. Alibaba Cloud Releases Qwen3-Max: A New Flagship Model with Over One Trillion Parameters

1. Released on September 24, with a parameter scale of over 1 trillion, it focuses on code and intelligent agent tasks. Officials claim that it has reached a leading level in multiple benchmarks.
2. The pre-training scale is approximately 36 trillion tokens (according to industry reports), and the multimodal Qwen3-Omni is launched simultaneously, targeting scenarios such as in-vehicle and AR/VR.
3. Combined with the upward signal of RMB 38 billion (about RMB 380 billion? See below) in AI infrastructure investment in the previous three years, this marks a milestone in the "model + computing power + application" combination.

2. Ali Tongyi Wanxiang Wan 2.5-Preview is online: native audio and video synchronization video generation

1. The preview will be open to developers on September 24, supporting text/image generation into videos, text generation into images, and conversational image editing.
2. The video can generate 1080p clips of about 5-10 seconds, and realize the joint generation of character dialogue, sound effects and pictures in synchronization.
3. Positioning as a "movie-level" controllable creation tool to fill the gaps in domestic multimodality in audio-visual alignment and lens motion stability.

III. "AI Plus" International Cooperation Initiative Released: Five Major Directions Aligned with the 2030 Agenda

1. It was announced on September 24, proposing to advance in five major directions: people's livelihood and welfare, scientific and technological progress, industrial application, cultural prosperity, and talent cultivation.
2. Emphasize the importance of conducting policy exchanges and practical cooperation based on the principles of mutual respect, mutual benefit and win-win results, while taking into account the inclusiveness of the Global South.
3. It is regarded as a public product in the field of AI provided by China to the international community, echoing the implementation of the domestic "Artificial Intelligence +" action opinions.

IV. Release of the "Consensus on Ethical Norms for Artificial Intelligence Serving Minors"

1. The results will be released at the Beijing Cultural Forum on September 23, focusing on seven principles including content assurance, privacy protection, transparency and credibility, and collaborative governance.
2. Industry associations will work with multiple parties to form a self-discipline consensus to connect with the previous generative AI safety guidelines for minors.
3. Propose clearer bottom lines and process requirements for product design and content review of education, social and content platforms.

5. Stargate Expansion: OpenAI, Oracle, and SoftBank Add Five New US Data Centers

1. On September 24, it was announced that new sites would be added: Shackelford, Texas; Milam, Texas; Lordstown, Ohio; Doña Ana, New Mexico; and one Midwest site.
2. The planned computing power is close to 7GW, and the cumulative investment scale will increase to more than US$400 billion in the next three years, with the goal of reaching US$500 billion/10GW by the end of 2025.
3. It is expected to drive approximately 25,000 on-site jobs and work in conjunction with the Abilene flagship and CoreWeave projects to alleviate the shortage of high-end AI computing power.

6. The EU inquired with multiple platforms about AI fraud management under DSA

1. On September 24, the European Union sent letters to Apple, Google, Microsoft, Booking, etc., asking them to explain their measures to combat financial fraud and AI deep fakes.
2. Alleging that related frauds cause losses of more than 4 billion euros each year, the platform is required to strengthen identification, removal and user protection.
3. Reflect the extension of DSA’s supervision in the governance of AI-driven fraud and false content.

7. UN High-Level Week Focuses on AI: Global Governance Forum Launches

1. On September 24, multilateral forums listed AI as a global challenge issue on par with climate change and epidemics.
2. The recently adopted resolution establishes a "Global Dialogue on AI Governance" and a 40-member independent scientific group, with meetings scheduled to be held in Geneva in 2026 and New York in 2027.
3. Although the mechanism is inclusive, its enforceability and binding force remain to be seen.

8. Meta Expands Llama to US Allies and International Organizations

1. On September 24, Llama expanded its availability to France, Germany, Italy, Japan, South Korea, EU institutions, NATO, etc.
2. Collaborative deployment with Microsoft, AWS, Oracle, Palantir, etc., for multimodal scenarios such as text, video, image, and audio.
3. Following the approval of the US government agencies on September 22, this is another move to increase the penetration rate of open source systems in government affairs and public sectors.

Frequently Asked Questions (Q&A)

Q: What is the core attraction of Qwen3-Max?

A: The parameter scale exceeds 1 trillion, emphasizing code and intelligent agent tasks; it is reported that about 36 trillion tokens are pre-trained, and the official said that many benchmarks are at the forefront, and the multimodal Qwen3-Omni is simultaneously launched for in-vehicle/AR, etc.

Q: To what extent will Stargate expand this time?

A: Five new sites have been added in the United States, with a planned computing capacity of nearly 7GW. The cumulative investment in three years will exceed US$400 billion, and the goal is to achieve US$500 billion/10GW by the end of 2025.

Q: What is the basis and focus of the EU’s inquiry?

A: According to the Digital Services Act (DSA), the focus is on how platforms can curb AI deep fakes and financial fraud; the EU said the related losses exceed 4 billion euros per year and asked the platforms to explain their governance plans.

Q: What areas does the “AI+” international cooperation initiative cover?

A: We will define five major directions: people's well-being, scientific and technological progress, industrial application, cultural prosperity, and talent cultivation, emphasize inclusive and pragmatic cooperation, and serve the UN 2030 Agenda.