Artificial intelligence research has taken another major step forward as Anthropic researchers have identified an internal activation space within its Claude AI models that appears to support complex reasoning and information processing. The company has named this internal structure J-space, describing it as a functional digital equivalent of the human brain’s global workspace.
According to Anthropic, the discovery provides new insights into how advanced AI systems process information internally. Researchers suggest that Claude’s architecture demonstrates several cognitive characteristics associated with conscious access in humans, including verbal reporting, directed modulation, internal reasoning, flexible generalization, and selective attention.
The findings highlight a growing effort among AI researchers to better understand what happens inside large language models while they generate responses, solve problems, and handle complex tasks. Rather than viewing AI systems as simple prediction engines, researchers are increasingly studying their internal mechanisms and how different parts of the models contribute to intelligent behavior.
The researchers describe J-space as a “silent workspace” where Claude carries out more deliberate and complex reasoning. Unlike traditional chain-of-thought methods, where an AI system produces visible reasoning steps in text form, J-space functions through internal neural activations that remain hidden from users.
Anthropic researchers found that Claude appears to use this internal workspace for tasks requiring deeper analysis, while simpler and more automatic processes are handled through other parts of the model. This separation between complex reasoning and routine processing resembles theories about how the human brain manages information through specialized systems.
One of the most notable aspects of the discovery is that J-space was not directly programmed into Claude. According to Anthropic, the workspace emerged naturally during the model’s training process, suggesting that advanced AI systems may develop internal structures that were not explicitly designed by engineers.
The research could have important implications for the future development of artificial intelligence. Understanding internal AI processes may help researchers improve model reliability, safety, transparency, and performance. It could also provide new methods for monitoring how AI systems make decisions and handle complex information.
Anthropic’s findings contribute to the broader scientific discussion surrounding machine intelligence and whether advanced AI systems develop internal representations that resemble aspects of human cognition. However, researchers emphasize that the discovery does not mean Claude is conscious or possesses human-like awareness.
Instead, J-space offers a new perspective on how AI models organize information and perform reasoning tasks. By studying these hidden internal mechanisms, scientists hope to build AI systems that are more understandable, trustworthy, and capable of assisting humans across a wide range of fields.
As artificial intelligence continues to evolve, discoveries like J-space demonstrate that the internal workings of modern AI models remain an important area of research. Future breakthroughs could further reveal how machines process knowledge, solve problems, and adapt to increasingly complex challenges.
