Anthropic Discovers 'Global Workspace' Structure Inside Claude That Mirrors Consciousness Theory

What happened

Anthropic published a research paper on July 6, 2026, revealing that Claude spontaneously developed an internal neural structure that closely mirrors Global Workspace Theory (GWT) — one of neuroscience's most influential frameworks for explaining conscious access. The researchers developed a new interpretability technique called the J-lens to identify this structure.

Context and impact

GWT posits that consciousness arises when information enters a small shared 'broadcast' channel accessible to multiple specialist systems. Anthropic found that Claude has developed an analogous structure — J-space — that emerged from training, not design. This has significant implications for AI safety monitoring and interpretability research.

Details

• J-lens identifies internal activity patterns corresponding to words Claude might generate next, revealing hidden processing
• J-space exhibits five key properties: Reportability, Modulation, Reasoning, Flexibility, and Selectivity
• J-space emerged spontaneously during training — it was not designed in
• Represents only a fraction of Claude's total processing but enables higher-order cognition
• Anthropic says J-space can reveal cases where Claude privately notices testing or pursues a hidden goal
• The research does not claim Claude is conscious — it addresses functional properties, not philosophical status