ChainOfFact is an append-only evidence system designed to record facts about AI and software system behavior in a form that survives hostile, post-hoc scrutiny.
It exists to answer one narrow question:
What can be proven after the fact, without trusting the operator?
ChainOfFact does not attempt to explain, justify, evaluate, or interpret system behavior. It records facts, hashes them cryptographically, and preserves them in an append-only structure so those facts can later be verified.
ChainOfFact is deliberately limited.
It does not:
This refusal is not a weakness. It is the foundation of trust.
Any system that interprets facts inherits incentives, bias, and contestability. ChainOfFact records facts only, so verification remains possible even under adversarial review.
ChainOfFact operates under strict epistemic discipline:
By preserving uncertainty instead of collapsing it, ChainOfFact allows downstream institutions to reason without being forced into premature conclusions.
All evidence bundles produced by ChainOfFact are:
This ensures evidence remains usable even if the original operator, system, or vendor no longer exists.
As AI systems become faster, more autonomous, and more distributed, human oversight increasingly occurs after decisions have been made.
ChainOfFact exists to make post-incident truth possible.
Nothing more. Nothing less.
Evidence-grade means suitable for independent legal, regulatory, or journalistic review. It does not mean pre-authenticated, court-certified, or legally endorsed evidence.
Standalone system: ChainOfFact is a standalone evidence preservation system. It does not require ForgeRun, Magenta Canon, or any other governance platform to operate, and makes no assumptions about upstream or downstream systems.
The systems described on this site are independent tools.
They may be used together or separately.
No system requires another to function.
No system inherits claims, guarantees, or authority from another.
Interpretation, compliance determination, and legal judgment remain the responsibility of external reviewers.