24-Hour AI News: OpenAI Parental Controls Launch, DeepSeek Price Cuts and Iterates

In the past 24 hours, domestic policies and implementation around "artificial intelligence +", computing power and communications have been frequently released. DeepSeek has iterated its model and lowered its price. Overseas, OpenAI launched parental control, US Medicare announced the AI review pilot schedule, and European chip research and development organization Imec completed AI-oriented personnel adjustments. Domestic and international dynamics remain balanced.

1. OpenAI launches ChatGPT parental controls to enhance youth safety

1. On September 29, OpenAI launched parental controls for users aged 13-18, supporting the association and management of accounts between parents and teenagers after mutual consent.
2. You can set silent periods, turn off voice/image generation, turn off memory and model training data usage, and other options.
3. When high-risk signals such as self-harm are detected and manually reviewed, the necessary level of safety notification will be sent to parents. The privacy does not include the specific content of the conversation.

2. DeepSeek releases V3.2-Exp "intermediate version" model and lowers API price

1. On September 29, DeepSeek released V3.2-Exp as a transitional version towards the next-generation architecture.
2. Introducing the DeepSeek Sparse Attention mechanism to focus on optimizing the training and inference efficiency of long texts.
3. The official announcement of a 50%+ reduction in API prices was also made. According to media reports, this measure was intended to verify the efficiency path while ensuring that the effects remained basically the same.

3. The time and scope of the US Medicare "WISeR" AI pilot program have been clarified.

1. The federal government has decided to launch a six-year "WISeR" pilot program in six states, including Washington and Texas, starting January 1, 2026.
2. AI participates in the prior authorization of some outpatient services. Any rejection requires manual review, and doctors can participate voluntarily.
3. The goal is to curb "low-value" medical expenses, and the outside world focuses on regulatory constraints on procedural fairness and the rate of erroneous injuries.

4. National Development and Reform Commission: Will formulate policy guidelines for the next generation of smart terminals and smart bodies

1. The press conference on September 29 revealed that principles and specific requirements will be put forward around technical support, application promotion and security governance.
2. Clearly accelerate standard research, promote open source sharing and scenario demonstration, and create a user-oriented evaluation system.
3. Positioning intelligent entities as direct carriers for the implementation of AI, emphasizing the demonstration application of "artificial intelligence +" in key areas such as education, medical care, and elderly care.

5. Imec appoints new CEO to advance AI-oriented chip R&D transformation

1. On September 29, Patrick Vandenameele will be appointed CEO of Belgian chip research and development center Imec, and the former CEO will be transferred to chairman.
2. The organization emphasizes collaboration with cloud hyperscale computing power providers, targeting the challenges of AI energy efficiency and process sustainability.
3. A sub-2nm test line is under construction, with part of the funding coming from the EU Chip Act, to serve the European deep technology ecosystem.

6. MIIT issues satellite mobile communications business license to China Mobile

1. On September 29, the Ministry of Industry and Information Technology issued a satellite mobile communications business license to China Mobile.
2. At this point, the three major operators all have the qualifications to carry out services such as mobile phone direct connection to satellite (S2C).
3. It will assist in scenarios such as emergency communications, maritime and remote area connections, and enrich the boundaries of AI IoT capabilities on the mass terminal side.

7. Cloudflare updates robots.txt "content signal policy" for the AI era

1. On September 29, Cloudflare launched refined signals that can distinguish search/AI-train/AI-crawl.
2. Provide publishers with clearer crawling and usage declaration channels to facilitate the distinction between retrieval, training and inference calls.
3. The industry believes that transparency has improved but enforceability remains controversial, and may evolve in parallel with industry licensing and litigation.

8. Nanning releases "Artificial Intelligence +" opening-up and cooperation measures

1. On September 29, new mechanisms such as the "AI Solutions International Exchange Market" were announced to deepen industry-university-research-application cooperation in ASEAN.
2. Released the AI cooperation vision for international sister cities and signed 33 investment projects, including 8 AI projects.
3. Establish a full-chain ecosystem of computing power, algorithms, and scenarios to promote the construction of cross-border smart logistics and trusted data zones.

9. National Cybersecurity Awareness Week: Large-scale model real-network public testing reveals 281 vulnerabilities

1. In September, a live network public test found 281 security vulnerabilities, more than 60% of which were specific to large models.
2. Typical risks include prompt injection, information leakage and data poisoning.
3. The industry is accelerating the construction of "AI against AI" security capabilities, and relevant assessment tools and solutions are being released intensively.

10. Maximor, founded by a former Microsoft executive, uses AI to replace Excel-style financial management and has secured $9 million in seed funding.

1. On September 29, Maximor announced financing and products, focusing on a financial AI agent network connected to ERP/accounting systems.
2. The goal is to shorten the monthly closing cycle, generate working papers and audit trails, and support GAAP/IFRS.
3. Targeting medium and large enterprise scenarios, emphasizing the closed loop of "human review + AI agency".

Frequently Asked Questions (Q&A)

Q: What exactly does OpenAI parental control cover?

A: It can be associated with a teenager's account after mutual consent; you can set silent periods, turn off voice/image generation, turn off memory and data training, turn on risk notifications, etc., and strengthen the filtering of sensitive content for teenagers by default.

Q: What is the milestone significance of DeepSeek V3.2-Exp?

A: This is an intermediate version towards the next-generation architecture. It introduces sparse attention to optimize the efficiency of long sequences, and with the reduction of API prices, it is expected to significantly reduce training and inference costs while achieving the same results.

Q: When will the Medicare “WISeR” pilot be launched and which states will it cover?

A: The plan will start on January 1, 2026 and last for 6 years, covering 6 states: Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. AI will be involved in some projects with prior authorization, and any rejection will require manual review.

Q: What does it mean that the Ministry of Industry and Information Technology issued a satellite mobile communications license to China Mobile?

A: The three major operators have all obtained satellite mobile licenses, and the commercialization of mobile phone direct connection to satellite (S2C) has been accelerated, which is beneficial to emergency communications and coverage in remote areas, and expands the space for "AI+connection" applications on the terminal side.

Q: What practical impact does Cloudflare’s robots.txt “content signal policy” have on media?

A: It improves the declarability of AI capture and usage (retrieval/training/inference can be distinguished), but due to the lack of enforcement, it still needs to comply with industry standards and legal paths. In the short term, it is more of a transparency and gaming tool.