24-hour AI news: The identification method has been implemented, and LongCat open source and OpenAI functions have been adjusted on the same day

24-hour AI news at a glance: The domestic AI content identification measures were officially implemented on September 1 and triggered platform announcements, and foreign OpenAI function adjustments and platform security governance were promoted simultaneously

In the past 24 hours (August 31-September 1, Beijing time), the deduplication check has been completed, removing the old topics in the channel article on August 30, and only including the latest progress. This issue of domestic and foreign news is balanced, focusing on regulatory implementation, model open source, platform governance and legal nodes.

1. The AI Content Identification Measures were officially implemented, and WeChat and Tencent Yuanbao issued announcements simultaneously

. From September 1, the "Measures for Identifying Artificial Generated Synthetic Content" officially came into effect, covering text, pictures, audio, video, etc.

1. On August 31, WeChat "Coral Security" and Tencent's "Yuanbao" announced respectively, requiring users to take the initiative to declare and keep the platform logo intact.
2. This move means that the platform needs to establish a full-chain traceability mechanism on the production and communication side, and violators may face disposal and accountability.

2. Meituan's open source large model LongCat-Flash, with total parameters of 560 billion

1. Adopt hybrid expert (MoE) architecture, with total parameters of 560 billion and dynamic activation of about 18.6B-31.3B.
2. The inference speed is nominally more than 100 tokens/s, supports 128k contexts, and is positioned as a "non-thinking" base model.
3. MIT protocol is open source and has been adapted to SGLang/vLLM for research and enterprise reuse.

3. China Telecom and Alibaba signed a strategic cooperation agreement

1. On August 31, the two sides signed a contract in Hangzhou and announced the deepening of cooperation.
2. Focus on cloud and AI infrastructure, new service industry e-commerce, and social value innovation.
3. It is expected to form a joint force in the field of computing power and industry applications to promote the coordinated development of the AI industry.

4. Manufacturing agent factories disclose efficiency improvement data

1. Factories have deployed more than ten types of agents, covering production scheduling, quality inspection, logistics and other links.
2. Achieve second-level response in multiple scenarios, with an average efficiency improvement of more than 80% and a production scheduling response speed increase of nearly 90%.
3. Provide cases for the combination of "agent + industrial data + robot" in the manufacturing industry.

5. OpenAI tests the "Thinking effort" selector

1. Provide ChatGPT with different levels of reasoning depth, and users can choose between speed and depth.
2. It is suitable for lightweight Q&A and complex analysis scenarios, which is convenient for cost and delay scheduling.
3. Reflect the idea of adjustable reasoning budget and provide more flexible runtime control for enterprises.

6. OpenAI's "Operator" will be discontinued on August 31, and the ability will be merged into the new ChatGPT Agent

1. Operator will be discontinued from August 31, and the function will be migrated to the new version of ChatGPT Agent.
2. The new agent supports retrieval, code execution, and connector calling, emphasizing the selection of autonomous tools.
3. Users need to migrate the workflow according to the official guidelines to avoid interruptions after September.

7. Meta restricts teenagers from accessing some AI characters and strengthens security governance

1. Due to doubts from the outside world, the platform announced that it will restrict some "provocative personality" characters from interacting with sensitive topics.
2. Adjustments involve topics such as self-harm and eating disorders, focusing on the safety of youth interaction.
3. This move reflects the tightening of platform governance under regulatory and public pressure.

8. New regulations for AI-related medical scenarios in Texas will take effect on September 1

1. The new regulations clarify the compliance boundaries of AI in medical and sensitive fields, emphasizing transparency and interpretability.
2. Stricter restrictions on the use of cross-border data and sensitive groups.
3. State-level laws provide a reference for the exploration of AI governance in the United States.

Frequently Asked Questions (Q&A)

Q: What are the key requirements of the AI Content Identification Method?

A: Explicit and implicit logos run parallel to cover all AI-generated synthetic content. The platform needs to review and prompt during the online process, and users must not delete or tamper with it.

Q: What are the highlights of Meituan LongCat-Flash?

A: The total parameters are 560 billion, MoE architecture, about 27B parameters are dynamically activated, the inference speed is fast, supports 128k contexts, and has been open sourced under the MIT protocol.

Q: What is the role of OpenAI's "thinking strength"?

A: Allowing users to choose the depth of reasoning, fast for light tasks, and deep for complex tasks, helps strike a balance between cost, latency, and quality.

Q: Why is Meta tightening the interaction between AI chatbots and minors?

A: Due to the risk of sensitive topics and public pressure, the platform temporarily restricts some roles and themes to enhance youth protection.