AI industry focuses on data quality for embodied intelligence and security for agents

Developments in the AI sector are shifting focus toward data quality for embodied intelligence and security frameworks for AI agents.

In the field of embodied AI, industry experts emphasize that data quality and rigorous acceptance systems are more critical than scale for improving model generalization and deployment stability. Key standards for physical AI data include trajectory error control within 1 centimeter, sub-millisecond temporal synchronization across multi-modal sensors, and label consistency. Companies like Meifeng Technology are implementing multi-dimensional inspection systems, such as ManiEval, to categorize data by quality rather than simply filtering it, creating a closed loop from collection to deployment.

Simultaneously, Ant Group is addressing the security requirements of the Agent economy. As AI evolves from answering questions to executing transactions and operating devices, security must move from a peripheral defense to a foundational infrastructure. Ant Group is developing the Agent Security and Trust Interconnection Protocol (ASL) to ensure identity, intent, and data trustworthiness during agent-to-agent collaborations. This includes establishing independent identity and permission systems for agents to prevent unauthorized actions or intent deviation during complex tasks.