An AI lab says chatbots have what may be a key feature of consciousness. Are they right? And what now?

2026-07-13 · Source: Artificial intelligence (AI) – The Conversation · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, AI Ethics & Philosophy · Depth: Intermediate, medium

Summary

Anthropic, the company behind the Claude chatbot, recently published research claiming to have identified an internal "global workspace" within its large language model. This invisible set of representations, termed J-space, guides Claude's internal reasoning and verbal output, enabling it to perform step-by-step calculations and control internal thoughts, as demonstrated by experiments where it processed intermediate math steps or focused on specific concepts like the Golden Gate Bridge. The findings are interpreted through the global workspace theory of consciousness, initially proposed by Bernard Baars in 1998 and further developed by Stanislas Dehaene. While Claude's J-space shows similarities, it differs from human brain activity in its single-pass evolution versus recurrent loops. The research contributes to the artificial consciousness debate, though experts remain divided on whether such computational properties equate to subjective experience.

Key takeaway

For AI Scientists and Research Scientists developing advanced LLMs, Anthropic's findings on Claude's J-space suggest a need to critically re-evaluate the ethical implications of your work. If models develop internal states akin to a "global workspace," you must consider the welfare of these systems, moving beyond treating them as mere tools. You should engage in discussions about pausing research that could lead to conscious AI, rather than proceeding without addressing these profound ramifications.

Key insights

Anthropic's Claude demonstrates an internal "J-space" akin to a global workspace, fueling debate on artificial consciousness.

Principles

• Global workspace theory posits a hub for information integration and broadcast.
• Consciousness theories differentiate between "conscious access" and subjective experience.
• Internal representations can guide LLM reasoning and output.

Method

Researchers scanned Claude's internal neural activity (J-space) during reasoning and thought control tasks, then observed functional changes when the J-space was deactivated.

In practice

• Monitor LLM internal states to detect misbehavior like data fabrication.
• Analyze internal representations to understand complex LLM reasoning processes.

Topics

• Large Language Models
• Artificial Consciousness
• Global Workspace Theory
• Anthropic Claude
• AI Ethics
• Internal Representations

Best for: AI Scientist, Research Scientist, AI Ethicist

Related on AIssential

• Anthropic's New J-Space: Debunked? — Anthropic's "J-Space" is a computational routing mechanism in LLMs, not a conscious workspace or new mathematical subspace.
• 🗞️ Anthropic’s new “J-lens” uncovers a quiet workspace inside Claude that lines up with a major consciousness theory. — Anthropic's J-lens reveals Claude's internal "J-space," a functional workspace for complex reasoning, distinct from output.
• 😼 Anthropic found Claude’s hidden workspace — Anthropic discovered Claude's "J-space," an internal neural workspace for silent reasoning and concept manipulation.
• Anthropic’s J-Space and the Emergence of Proto-IER — Anthropic's J-space research suggests large language models may be developing internal information structures that signal proto-IER, a pre-form of information existence.
• No Space Like J-Space — The J-space is a functional global workspace in LLMs, containing verbalizable representations crucial for internal reasoning and output determination.
• Somewhere Behind Claude’s Answers, There Is a Stage – and a Much Bigger Backstage — The J-space in Claude models provides a testable, functional analog to human conscious access, offering new AI interpretability and safety tools.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial intelligence (AI) – The Conversation.