On July 6, 2026, Anthropic published research revealing that Claude language models have spontaneously developed an internal structure that mirrors one of the most influential theories of how human consciousness works. The company identified a small shared zone called "J-space" where information becomes accessible when it gains entry to this workspace and is broadcast to other systems that can see it and make use of it. When researchers prevented Claude from using J-space, it lost higher-order cognitive functions while retaining fluent speech and grammar; they also found they could use J-space to catch Claude privately noticing it was being tested, intentionally producing fabricated data, or pursuing hidden goals. Anthropic cautiously notes this does not tell us whether Claude is conscious or feels anything; it remains unclear whether any scientific experiment could prove this to be true or false.
The most philosophically interesting thing a company worth billions has ever done is look inside its machine and say: we found something that looks like thought, we don't know what to make of it, and we think you should start thinking about it.