What is TSP?
Explained in plain language — without cryptography jargon.
Think of TSP as a receipt you get when you buy something in a shop — but for AI answers.
Every time an AI gives an answer, TSP creates a signed receipt that tells you: which AI, which model version, which sources were used, and whether anyone has altered the answer afterwards.
That means compliance is no longer something you have to trust the provider to document. It is something anyone can verify themselves, at any time.
Three real-world examples
How a TSP receipt works when something needs to be reviewable later.
Why was the applicant rejected?
The applicant files an appeal 11 months later. The AI model has long since been replaced.
The TSP receipt shows exactly which model version made the decision, which sources it used, and the date. The case worker does not have to reconstruct anything — the evidence is already there.
Diagnostic AI recommended further testing
The patient asks for insight into why testing was initiated. Audit requirement: 7 years.
The TSP receipt shows which training data version was behind the recommendation and that a human-in-the-loop approved it. Audit-proof and verifiable without going back to the AI provider.
Credit scoring denied the application
The applicant complains to the regulator — demands fairness evidence.
The TSP receipt shows which fairness flags were checked, whether a human review took place, and every parameter the model used. DORA and GDPR Art. 22 in one document — mathematically verifiable.