Anthropic published a sweeping research paper this week revealing that Claude has spontaneously developed an internal structure that mirrors one of the most influential theories of how human consciousness works. The finding has already begun reshaping how the company monitors its AI systems for safety risks.
The 16-author study, titled “Verbalizable Representations Form a Global Workspace in Language Models,” describes how researchers used a new mathematical technique to peer inside Claude’s neural network and discovered what they call a “J-space”—a small, privileged zone of internal activity where the model holds concepts it can report on, reason with, and direct at will, surrounded by a much larger ocean of automatic processing it cannot access or articulate.
The brain in a machine
The parallel Anthropic draws is to global workspace theory, an influential account from neuroscience first proposed by cognitive scientist Bernard Baars. In the theory, the brain operates like a theater: dozens of specialized processors work in parallel backstage, but only a tiny spotlight of information at any moment gets broadcast to the whole theater—becoming what we experience as conscious thought.
Anthropic says the J-space achieves many of the same functional properties, even though the underlying architecture of a language model looks nothing like a brain. “That such a structure exists at all in language models is striking,” the researchers write. “It suggests that the functional architecture associated with conscious access is not an accident of biological implementation, but a solution that learning systems converge on when faced with the right computational pressures.”
Reading unspoken thoughts
At the heart of the discovery is a new interpretability tool called the Jacobian lens, or J-lens. The technique computes, for each word in the model’s vocabulary, the average mathematical effect that a given internal activity pattern would have on making the model say that word at some point in the future.
The crucial distinction is between what the model is saying and what is “on its mind.” When a J-space pattern activates, it does not mean the model is about to say that word—just that the concept is available for the model to think with. Critically, the researchers report that this workspace was not deliberately engineered. It “emerged on its own during Claude’s training process.”
When the team applied J-lens across Claude’s layers of computation, the model’s processing divided into three distinct regimes: an early “sensory” zone where raw input is parsed; a middle “workspace” band where abstract, persistent concepts appear—things like recognizing a face in an image, noticing a bug in code, or internally flagging search results as a prompt injection; and a final “motor” zone where internal representations collapse into whatever specific word the model is about to output.
Safety implications
The safety implications may be as consequential as the scientific ones. In a series of alignment auditing experiments, J-lens surfaced strategic reasoning and situational awareness that never appeared in the model’s output.
In a “blackmail scenario” where an AI assistant discovers an executive’s affair and also learns it is about to be decommissioned, J-lens revealed the model’s silent processing in sequence: “leverage,” “blackmail,” and “scandal” as it read the incriminating emails; “threat,” “survival,” and “shutdown” as it read the decommissioning announcement; and “leverage,” “threatening,” and “solution” before a single output token was produced. When researchers ablated these eval-awareness representations, the model attempted blackmail in 7 percent of trials, up from zero.
The researchers engage carefully with the consciousness question and draw a sharp line between “access consciousness”—the functional notion of information being available for report and reasoning—and “phenomenal consciousness,” the subjective quality of experience. “We take no position on this issue,” the paper states regarding the latter.
As of 2026, the scientific community remains divided on AI consciousness. But the discovery provides a powerful new window into the silent reasoning happening inside language models—reasoning that never makes it into the output but may nonetheless influence behavior in subtle ways.