AI Daily: State Administration for Radio, Film, and Television Crackdown on Chaos in AI Animation; 360 Releases Big Model Security White Paper; Baidu Launches Xiaodu AI Glasses Pro

Welcome to the "AI Daily" section! Here is your guide to exploring the world of artificial intelligence every day. Every day, we present you with the latest content in the AI field, focusing on developers to help you understand technical trends and innovative AI product applications.

Fresh AI products Click to learn more:https://app.aibase.com/zh

1. State Administration of Radio, Film and Television Cracks Down on AI Animation Chaos! From March 2026, Funny Short Plays and AI-Generated Content Will Be Fully Reviewed

The State Administration of Radio, Film and Television announced that it will launch a special governance campaign targeting AI-generated content nationwide before March 2026, explicitly incorporating AIGC works into a classified and layered review system to combat vulgar and distorted values, promoting the industry's transition toward high-quality development.

AiBase Summary:

🛑 Before March 2026, AI-generated animation short videos must pass review and obtain a filing number before going online.

🚫 Prohibited from using scandalous figures' images or voices for creation, forbidden to mock children's IP and alter classic animation dialogues.

💡 Regulation aims to shift the industry from "traffic-driven" to "quality-driven," creating space for high-quality content.

2. Wuzhen Summit Announces Major Release! 360 Launches the "Large Model Security White Paper", First Proposing a "Full-Chain AI Security Defense" to Address New Threats in the Intelligent Era

The article points out that as large models are increasingly applied in key areas such as finance and government affairs, the security risks they bring are becoming more prominent. The "Large Model Security White Paper" released by 360 Digital Security Group systematically reveals five core risks throughout the entire lifecycle of large models and proposes a 'external security + platform-native security' dual governance framework to provide a multi-layered defense barrier for the high-risk AI era.

AiBase Summary:

🧠 The white paper reveals five core risks across the entire lifecycle of large models, covering infrastructure layer, content layer, data and knowledge base layer, intelligent agent layer, and user end layer.

🛡️ Proposes a 'external security + platform-native security' dual governance framework, achieving security protection from external monitoring to internal embedding.

🌐 Emphasizes building an open, collaborative, and trustworthy AI security ecosystem, promoting standard setting, threat intelligence sharing, and open-source tool co-development.

3. Baidu Launches Xiaodu AI Glasses Pro, Price 2299 Yuan, Supports Multiple Smart Features

Baidu's Xiaodu AI Glasses Pro is a device that combines smart features with stylish design, offering multiple styles and lens options, along with powerful hardware configurations and AI functions, providing users with a convenient experience.

AiBase Summary:

📸 The glasses are priced at 2299 yuan, supporting 4K photography and AI translation among other smart functions.

🔋 Equipped with a 3000mAh charging box, the glasses have a battery life of up to 68 hours, suitable for long-term use.

🎤 Uses a five-microphone array, featuring high-quality sound and anti-leakage design, enhancing user experience.

4. StepFun AI Launches Open-Source Audio Editing Model Step-Audio-EditX, Providing a New Audio Editing Experience

StepFun AI has released the open-source audio editing model Step-Audio-EditX, which transforms audio editing tasks into text-based token operations, making voice editing more direct and controllable. The model uses a dual codebook tokenizer and large margin learning methods, showing excellent performance in emotional and stylistic editing, and introduces new benchmarks to improve audio quality evaluation.

AiBase Summary:

🎤 StepFun AI launches the open-source audio editing model Step-Audio-EditX, making audio editing simpler.

📈 This model uses large margin learning to improve the accuracy of emotional and stylistic editing.

🔍 Introduces Step-Audio-Edit-Test benchmark, significantly improving audio quality evaluation.

Details: https://arxiv.org/abs/2511.03601

5. Grok Now Supports Text-to-Video Generation

Grok Imagine launched a new feature that supports text-to-video generation, allowing users to input descriptions and generate video clips with background sound effects, dynamic camera movements, and professional quality within 17 seconds. This feature not only improves video creation efficiency but also achieves human-computer co-creation through a multimodal interaction loop.

AiBase Summary:

✨ Supports text-to-video generation, completing video production in 17 seconds.

🎨 Offers multiple style switches and creative modes to meet diverse creative needs.

🔄 Achieves human-computer co-creation, supporting real-time adjustment of prompts to optimize video results.

6. Study Finds Google AI Model Veo-3 Can Generate Realistic Surgical Videos but Lacks Medical Logic Understanding

A study found that the Google AI model Veo-3 can generate realistic surgical videos but lacks medical logic understanding. Although its visual effects are astonishing, there are significant shortcomings in medical operation understanding, especially in instrument usage, tissue response, and surgical logic.

AiBase Summary:

🧠 The Veo-3 model can generate realistic surgical videos but lacks medical logic understanding.

🔍 Over 93% of errors stem from medical logic issues, seriously affecting video accuracy.

📈 The research team plans to open-source the dataset to promote progress in AI's medical understanding.

7. Alibaba's Qwen3-Max-Thinking Wins First Place in Global Math Competition, Challenging OpenAI's Leadership

Alibaba's Qwen3-Max-Thinking performed outstandingly in the global math competition, demonstrating its strong reasoning and problem-solving abilities and achieving significant investment returns in cryptocurrency trading, challenging OpenAI's leadership.

AiBase Summary:

🧠 Alibaba's Qwen3-Max-Thinking achieved a perfect score in international math competitions, challenging OpenAI.

🏆 This model has over 1 trillion parameters and is the first Chinese AI to achieve 100% accuracy in AIME and HMMT.

📈 Qwen3-Max performed well in real-world cryptocurrency trading, achieving a 22.3% investment return.

8. OpenAI Launches a Lighter GPT-5 Codex Mini Model

OpenAI launched the lighter GPT-5 Codex Mini model, optimizing the underlying structure of Codex, enhancing the service experience, and providing developers with more choices and a more efficient workflow.

AiBase Summary:

🧠 The GPT-5 Codex Mini model focuses on providing more efficient and cost-effective code generation capabilities, offering developers more flexible choices.

🚀 OpenAI has increased the rate limits for ChatGPT Plus, Business, and Edu users, and provides priority processing for Pro and Enterprise users.

🛠️ GPT-5 Codex Mini supports command-line interface (CLI) and integration with integrated development environments (IDEs), and the API interface will be available soon.