AI that affects people should come with a receipt.
This is intentionally plain. A receipt does not claim perfection. It tells people the minimum facts they should not have to beg for: what the AI is doing, who owns it, what data it touches, how a person can intervene, and what evidence exists when something goes wrong.
PAI maps the receipt back to the Production Safety Framework and PAI-8 so small organisations can start with transparency, while serious deployments have a path into stronger evidence, Lab review, and formal assurance.
For business owners
Publish a receipt before customers ask for one. It shows maturity without pretending a lightweight disclosure is a full audit.
Generate a receipt ->For MSPs and consultants
Run a client session, capture the workflow, and leave behind a public transparency artifact that opens the deeper control conversation.
Open the MSP pack ->For governments and regulators
Use the receipt as a simple public expectation: AI systems that affect people should have ownership, boundaries, and an escalation route.
Read PAI-8 ->For customers and workers
Ask for the receipt. It is a direct way to request basic clarity without needing to understand the entire technical stack.
View registry ->Useful first, citeable second, institutional by design.
Generate
A business, MSP, or consultant creates a receipt in minutes.
Publish
The receipt gets a stable public URL and registry entry.
Share
Customers, staff, partners, and local communities can ask others to match the standard.
Improve
The receipt points to concrete PSF actions, not vague promises.