大湾区首家通用具身智能大模型完成备案

Summary 1: The most noteworthy point is that the first general-purpose embodied AI foundation model in the Greater Bay Area has completed filing, and the model has entered real-world scenarios.

1. X Square Robot's self-developed GreatWall general-purpose embodied AI model passed Guangdong's generative AI service filing, becoming the first general-purpose embodied AI foundation model in the Greater Bay Area to complete filing.

2. Before filing, it completed a five-month safety assessment under national standards, conducted more than 200,000 spot checks on training data and generated content, completed 27,148 high-risk refusal tests, and established a safety safeguard mechanism covering data input through service operation.

Summary 2: Model capabilities and application scenarios are the other key takeaways.

1. The GreatWall series includes the world unified model WALL-B, the world model WALL-WM, and the VLA model WALL-OSS.

2. WALL-B uses a pretraining, end-to-end, and general-purpose embodied AI foundation model technical route, integrating vision, language, touch, action, and physical prediction, enabling robots to understand the relationships among environments, objects, and tasks and to perceive, judge, and operate autonomously.

3. The model can be reused across embodiments, tasks, and scenarios and has been applied at scale in real-world scenarios such as home services and logistics sorting.

Summary 1: The core information for brands is the brand trust and product endorsement created by compliant filing.

1. GreatWall has become the first general-purpose embodied AI foundation model in the Greater Bay Area to complete filing, marking new progress in model compliance, productization, and large-scale application.

2. Before filing, it completed a five-month safety assessment covering training data, generated content, intellectual property, personal information protection, accuracy, reliability, and risk interception, and established a safety safeguard mechanism from data input to service operation. This can serve as a safety and compliance selling point in brand communications.

Summary 2: On product R&D and consumer trends, the article points to embodied AI entering real-world scenarios.

1. The GreatWall series covers WALL-B, WALL-WM, and WALL-OSS. WALL-B integrates vision, language, touch, action, and physical prediction and can be reused across embodiments, tasks, and scenarios.

2. Large-scale application has already occurred in real-world scenarios such as home services and logistics sorting, indicating that home and logistics are demand scenarios worth watching for embodied AI brands; safety, compliance, and reliability are clearly defined as the foundation for scaling.

3. The article says little about pricing and price competition, focusing instead on how compliance and safety governance support brands entering real-world scenarios. Events such as the Guangdong filing model launch conference provide a platform for technical exchange and industry cooperation for compliant models, which can be viewed as an observation point for brand channels and industry partnerships.

Summary 1: At the policy level, Guangdong's generative AI service filing information has been updated, and the compliance window and industry environment merit attention.

1. As of November 11, Guangdong added three generative AI services that completed filing, bringing the cumulative number to 230 and keeping the province among the national leaders. Guangdong is one of the most active regions in China's generative AI industry.

2. X Square Robot's GreatWall general-purpose embodied AI model completed filing, becoming the first such model in the Greater Bay Area to do so, indicating progress in the compliance of general-purpose embodied AI foundation models.

Summary 2: Opportunities, risks, and lessons center on real-world deployment and safety compliance.

1. Opportunity signal: WALL-B can be reused across embodiments, tasks, and scenarios and has already been applied at scale in real-world scenarios such as home services and logistics sorting. Embodied AI must ultimately enter real environments such as homes, logistics, and industry.

2. Risk signal: safety, compliance, and reliability are not add-ons but the foundation for scaling. Models must complete assessments covering training data, generated content, intellectual property, personal information protection, accuracy, reliability, and risk interception.

3. Lessons: the five-month safety assessment, more than 200,000 spot checks, 27,148 high-risk refusal tests, Guangdong's enterprise-close service mechanisms, and the technical exchange and industry cooperation platform created by the filing model launch conference.

Summary 1: For factories, the article's product design and production needs point to combining embodied AI models with real work scenarios.

1. The GreatWall series includes the world unified model WALL-B, the world model WALL-WM, and the VLA model WALL-OSS. WALL-B uses a pretraining, end-to-end, and general-purpose embodied AI foundation model technical route, integrating vision, language, touch, action, and physical prediction into a unified model.

2. The model enables robots to understand the relationships among environments, objects, and tasks, autonomously complete perception, judgment, and operation, and offers reuse across embodiments, tasks, and scenarios. This creates requirements for multimodal perception, action execution, and scenario adaptation in product design.

Summary 2: Business opportunities and digitalization implications come from scenario deployment and compliant productization.

1. Business opportunities: the model has been applied at scale in real-world scenarios such as home services and logistics sorting. Embodied AI will also enter real environments such as homes, logistics, and industry; factories can watch for opportunities in logistics sorting, home-service-related products, and production lines.

2. Digitalization and e-commerce implications: although the article does not directly discuss e-commerce, the model's compliance, productization, and large-scale application, along with the safety safeguard mechanism from data input to service operation, show that digital products entering real-world scenarios must first address safety, compliance, and reliability.

Summary 1: On industry trends, general-purpose embodied AI foundation models are entering the stage of compliance filing and large-scale real-world application.

1. X Square Robot's GreatWall has become the first general-purpose embodied AI foundation model in the Greater Bay Area to complete filing, marking new progress in model compliance, productization, and large-scale application. Guangdong's cumulative filings have reached 230, ranking among the national leaders, with a significant industry clustering effect.

2. The GreatWall series covers WALL-B, WALL-WM, and WALL-OSS. WALL-B uses a pretraining, end-to-end, and general-purpose embodied AI foundation model technical route, integrating vision, language, touch, action, and physical prediction.

Summary 2: Customer pain points and solutions center on safety compliance and scenario reuse.

1. Pain points: when embodied AI enters real environments such as homes, logistics, and industry, safety, compliance, and reliability are the foundation for scaling, involving training data, generated content, intellectual property, personal information protection, accuracy, reliability, and risk interception.

2. Solutions: before filing, it completed a five-month safety assessment, more than 200,000 spot checks on training data and generated content, and 27,148 high-risk refusal tests, and established a safety safeguard mechanism from data input to service operation. The model can be reused across embodiments, tasks, and scenarios and has been applied at scale in home services and logistics sorting.

Summary 1: The article shows that demand from platform ecosystems for compliant large models centers on filing services, showcase and exchange, and industry cooperation.

1. The Guangdong cyberspace administration released a new batch of generative AI service filing information. As of November 11, Guangdong had added three services that completed filing, bringing the cumulative total to 230 and keeping the province among the national leaders, with a significant industry clustering effect.

2. Guangdong continues to improve its generative AI governance and service system and uses events such as the filing model launch conference to showcase phased results, creating a platform for technical exchange and industry cooperation for compliant models.

Summary 2: Platform operations and trend-risk avoidance should focus on compliance thresholds and safety governance.

1. Platform investment promotion and operations can be organized around compliant models: filing information disclosure, results showcases, technical exchange, and industry cooperation can help form a cluster of compliant models and facilitate cooperation.

2. Trend-risk avoidance points: models must complete safety assessments covering training data, generated content, intellectual property, personal information protection, accuracy, reliability, and risk interception. GreatWall underwent a five-month assessment before filing, more than 200,000 spot checks, and 27,148 high-risk refusal tests, and established a safety safeguard mechanism from data input to service operation.

3. Platforms also need to focus on safety baselines, product standards, and reliability in real-world applications.

Summary 1: The new industry development is that the first general-purpose embodied AI foundation model in the Greater Bay Area has completed filing, while Guangdong's filing scale continues to lead.

1. X Square Robot's self-developed GreatWall general-purpose embodied AI model passed Guangdong's generative AI service filing, becoming the first general-purpose embodied AI foundation model in the Greater Bay Area to complete filing.

2. As of November 11, Guangdong had added three generative AI services that completed filing, bringing the cumulative total to 230 and keeping the province among the national leaders, with a significant industry clustering effect.

Summary 2: New questions, policy implications, and business models are worth studying.

1. New questions: when embodied AI enters real environments such as homes, logistics, and industry, safety, compliance, and reliability become the foundation for scaling, involving training data, generated content, intellectual property, personal information protection, accuracy, reliability, and risk interception.

2. Policy and regulatory implications: Guangdong is improving filing rules and enterprise-close service mechanisms, providing clear and stable expectations for technological innovation. Through events such as the filing model launch conference, it builds a platform for technical exchange and industry cooperation, promoting coordinated development of innovation and safety governance.

3. Business model: the GreatWall series includes WALL-B, WALL-WM, and WALL-OSS. WALL-B integrates vision, language, touch, action, and physical prediction, can be reused across embodiments, tasks, and scenarios, and has been applied at scale in home services and logistics sorting.