The architecture of artificial intelligence is undergoing a structural shift. In the span of a single week, two developments have made the change visible: the launch of a feature that lets the AI model Claude join group conversations inside the workplace platform Slack, and a $2.8 billion valuation for the Chinese start-up X Square Robot, whose machines learn from real-world interaction rather than pre-scripted commands. Viewed together, they mark a move away from AI as a solitary digital assistant and toward systems that operate in teams, inhabit physical bodies, and manage complex, unpredictable environments.

This reorientation is technical as much as conceptual. X Square Robot’s WALL-B foundation model fuses vision, language and movement into a single system, allowing robots to understand instructions, recognise objects and react to changing surroundings without separate programming for each action. The company has deployed its machines in occupied homes, elderly care facilities and warehouses, building a data pipeline in which every completed task feeds back into the model. Meanwhile, the Claude Tag feature, developed by Anthropic and Salesforce, enables an AI to read a group’s message history, divide complex projects into stages, schedule them and execute steps autonomously—all while remaining visible to the entire team. In both cases, the AI is no longer a private interlocutor but a participant in shared physical or digital spaces.

This evolution is unfolding against a backdrop of intensifying great-power competition. A new global arms race, estimated at $2 trillion, is centred on hypersonic missiles, military AI, space weapons and autonomous drones. Washington is planning a $1.5 trillion defence budget and has already used AI in operational planning, while Beijing’s effective military spending may reach $500 billion, with a focus on fusing space, cyber and autonomous systems. Moscow, despite financial strain, has tripled its military budget and tested hypersonic weapons in combat. European states, collectively spending around $600 billion, are developing sovereign AI and secure space communications. In parallel, a quieter struggle is playing out over the intellectual property of AI models themselves. American firms allege that Chinese competitors are using model distillation—a technique in which a large model trains a smaller one—to extract knowledge from systems like ChatGPT and Claude without authorisation, narrowing the capability gap to as little as 12 to 24 months.

Analysts in Beijing and Washington note that China’s AI trajectory is distinct. Rather than concentrating on chatbots or artificial general intelligence, Chinese investment is flowing into systems for real-time coordination: traffic management, predictive logistics, digital twins of entire cities and intelligent manufacturing. This approach, which some observers link to older Chinese traditions of thinking about change and interdependence, treats AI as infrastructure for governing movement and absorbing complexity. The immediate milestones to watch are the commercial scaling of X Square Robot’s cleaning and care services, the enterprise uptake of group-based AI tools, and any regulatory or export-control responses from Washington to the distillation allegations.