RECEIPTS

WHAT CAN WE ACTUALLY SAY?

Supported in the current internal record

WantWan's source work has documented recurring cases where an AI system's account of remembering, checking, executing or correcting did not fully match the underlying record. Related events were grouped in the internal audit; the pattern still recurred in the available record.

Still open

The current record does not establish population prevalence, comparative rarity, superiority over professional evaluators, a validated hit-rate or false-positive rate, transfer to unfamiliar AI systems, validated market pricing, a mechanism for the observed behaviour, or machine consciousness.

How a public receipt should work

Question → source/setup → what the AI said → what happened → correction/intervention → did it survive? → current interpretation → limitation → why it matters.

Public summaries are deliberately less revealing than the private source corpus. Evidence access is controlled so that verification does not automatically disclose private material, proprietary methodology or unrestricted reuse rights.