Why it’s still open
A 2026 study reports a small, reportable, broadcast set of representations in Claude that behaves like a global workspace, and leading workspace theorists have responded with both parallels and caveats. What this shows about consciousness, as opposed to access and report, is exactly what remains disputed.
What would count as an answer
Replication across models and architectures, the further tests workspace theorists propose, and an argument connecting a functional workspace to phenomenal consciousness that survives the usual objections.
Papers that tackle it
- Chalmers 2023 - Could a large language model be conscious — Weighs reasons for and against LLM consciousness; names missing features such as recurrence, a global workspace and unified agency as obstacles for the models of the time, while expecting successors may overcome them.
- Gurnee Sofroniew et al 2026 - A global workspace in language models — Introduces the Jacobian lens and finds a small set of verbalizable representations in middle layers that behave like a global workspace: reportable, controllable, used in reasoning and broadcast widely, while routine processing stays outside it.
- Dehaene Naccache 2026 - Does Claude possess a conscious global workspace — Two global-workspace theorists compare the reported J-space with the human global neuronal workspace, set out where the parallel holds and where it does not, and propose further tests.
- Goldstein Kirk-Giannini 2024 - A case for AI consciousness — Argues that if global workspace theory is correct, language agents could easily be made phenomenally conscious, and sets out a method for applying scientific theories of consciousness to AI.
- Butlin Long et al 2023 - Consciousness in AI — Derives indicator properties from leading theories of consciousness and checks AI systems against them; concludes no current system is a strong candidate but finds no obvious barrier to building one.
- Butlin et al 2026 - Eleos commentary on the global workspace — Strong evidence for access, weaker for a unified workspace.
Methods in play
Who’s closest
Placed by the editor, v1 — corrections welcome.
- Wes Gurnee
- Nicholas Sofroniew
- Jack Lindsey
- Stanislas Dehaene
- Simon Goldstein
- Cameron Domenico Kirk-Giannini
- David Chalmers
Related questions
Know a paper that belongs here, or think this is framed wrongly? Suggest a paper or a correction (forms on the Atlas home).