Anthropic's research team has made a breakthrough by discovering a previously unknown internal structure within its language model Claude. This mechanism, named "J-space," functions as a common workspace that various components of the model access. It is important to emphasize: engineers did not build this architecture into the design during development — it emerged spontaneously during training.
What is J-space and how does it work?
At its core, J-space is an internal virtual space where Claude collects and transmits key data throughout the model. It can be thought of as a kind of "board" inside the artificial intelligence. When the model answers a question or solves a task, critically important information appears in J-space so that different parts of the model can interact with it. Scientists identified this structure using a special tool called "J-lens," which allows observing the movement of information during task execution.
Notably, J-space largely mirrors the concept of a "global workspace" in human thinking. When a person hears a question, recalls a needed fact, and decides how to answer, their brain brings all necessary data together in one place. Claude works in a similar way. Moreover, researchers demonstrated that the model can describe the contents of J-space upon request and even modify it if asked. Manual intervention in J-space directly affected the model's responses and behavior.
Why this changes the game for AI safety
This discovery has enormous implications for the safety and interpretability of artificial intelligence. The ability to track J-space activity opens the way to identifying hidden motives in model behavior. We will be able to more effectively detect moments when the system begins to operate unreliably. In particular, monitoring J-space allows seeing attempts at prompt injections — the insertion of malicious instructions into a query to hijack control of the model's response.
Of course, the capabilities are still limited. Most of Claude's information processing still occurs outside J-space. Nevertheless, Anthropic has already published the source code for the J-lens implementation and released a demo version on Neuronpedia, inviting the research community to verify the results in practice.
This research builds on a series of earlier works. In October 2025, Anthropic published a report on emerging introspective "awareness" in models, and in April launched initiatives to study their well-being. An important caveat: the company does not claim that Claude possesses consciousness or subjective experience. The term "consciously accessible" information is borrowed from cognitive science and does not imply the presence of actual consciousness.
Comment from Cryptalist expert: The discovery of J-space is not just an academic curiosity. It is a step toward creating an "X-ray machine" for neural networks. The ability to look inside the "black box" of decision-making is the key to building truly safe and controllable AI systems, which is critically important for their integration into financial and management processes.