Anthropic claims Claude has global workspace, authenticity of consciousness and safety control spark controversy

On July 6, 2026, Anthropic released a study stating that Claude has a "J-space" internal workspace similar to human consciousness, capable of unoutputted reasoning and self-awareness.

The study comes from Anthropic's Interpretability team, whose mission is to understand the internal working mechanisms of large language models in order to lay the foundation for AI safety. The report points out that this workspace allows the model to engage in internal thinking without generating outputs, similar to the global workspace theory of human consciousness. The official blog also lists other research, including the Economic Index report from June 26, 2026, and alignment research from May 8, 2026, showing that Interpretability work is part of Anthropic's overall safety strategy.

Mechanism Breakdown

From a technical logic perspective, the emergence of a global workspace stems from the representation capability of internal states that emerge after model scaling. Anthropic's research shows that Claude can maintain unoutputted reasoning chains across multiple interactions, echoing earlier alignment team efforts to reduce agentic dysfunction. On the commercial level, this finding may serve Anthropic's long-term investment in model interpretability, as the Interpretability team explicitly views understanding internal mechanisms as a prerequisite for positive AI impact.

Industry Impact

Regarding the competitive landscape, this study strengthens Anthropic's positioning in the interpretability field, potentially attracting safety-conscious developers to its API. For enterprise users, the concept of an internal workspace may be used to audit AI decision-making processes, but it could also become a reason to restrict model autonomy. For upstream and downstream developers, the research implies that future tools need to support monitoring of unoutputted states, which will increase integration complexity.

Strategic Judgment

Based on existing facts, Anthropic will continue to publish more Interpretability results to support its safety narrative, while opponents may cite this research to push for stricter regulatory frameworks.