Oracle Ethics is a research prototype exploring exactly that. Every answer it generates comes with a "deception probability" score and a cryptographic hash, so you can audit its honesty. What do you think is verifiable transparency the future of trustworthy AI?