ATA-1: A Public Draft for AI Money-Movement Controls
ATA-1 is a public draft of controls for how AI systems should be authorized, limited, and logged when they move money. Its README and ATA-1.md list it as v0.1, organized into 30 controls across 7 domains, released under a CC BY 4.0 license, with Ethan Cho named as maintainer.
The starting question
If an AI moves money incorrectly, can anyone later verify who approved what and on what basis?
Approach
The document structures its 30 controls across 7 domains and outlines proposed certification levels. Those levels are the author's own proposed draft structure; the underlying evidence specification, quantitative thresholds, and mappings to other standards are marked as open items in the published documents.
Artifact and current status
The README and ATA-1.md are readable directly in the public repository at github.com/Barac9492/ata-1; this reflects the document state observed on September 16, 2026. A permission-approval demo on the author's homepage is a separate, deterministic illustration of the idea, not an implementation of the ATA-1 standard, and it makes no real payments or AI calls.
Limits and what is not verified
The certification levels described are the author’s own proposal, not an adopted or insurer-approved certification scheme. The public materials do not establish insurer or institutional adoption, or a real assurance deployment; that is a limit of what the public record currently shows, not a claim that no such deployment exists.