AI agents can act for you. AT1C makes sure they only act with your permission — and proves it, cryptographically, every time.
The evidence layer for human control of AI
"Don't trust us. Verify the receipt."
AI tools make decisions. Platforms control your access. Your agents act — and there is no cryptographic record that you said yes.
AI agents lack intention, so liability can only attach to the human who approved the action. But only if there is evidence of what they approved.
AI agents lack intention, so liability can only attach to the human who approved the action — but only if there is evidence of what they approved.
— Ayres & Balkin, Yale Law School / University of Chicago Law Review, 2026Your agent acts. There is no proof you approved it. Under the EU AI Act, that is a liability. In court, that is no evidence.
Every action is backed by a cryptographically signed proof of your approval — independently verifiable by any auditor, regulator, or counterparty.
Agent asks permission for a specific action — scoped, named, bounded.
You explicitly grant or deny it. Your key. Your decision. No platform in between.
Signed receipt generated — timestamped, cryptographic, your permanent evidence.
Any party verifies independently — no trust needed, no intermediary required.
No wallets. No seed phrases. No complexity. Register in minutes.
AT1C works for anyone whose AI agent takes actions on their behalf.
Own a verifiable record of every action your personal AI takes. Approve, revoke, and audit — with no platform in between.
Manage fleets of agents across business processes. EU AI Act compliant from day one. Free, open source, wraps your existing stack.
High-risk AI sectors where every action needs a verifiable human approval record — for regulators, auditors, and counterparties.
ML-DSA-65 signing, receipt generation, verification. MIT licensed. Drop in and go.
Large companies spend €400k–€2M per year on AI Act compliance infrastructure. AT1C gives SMEs the same verifiable accountability in a day.
| Article | Requirement | AT1C delivers |
|---|---|---|
| Art. 9 | Risk management | Approval checkpoint at every action |
| Art. 13 | Transparency | Request names action explicitly before approval |
| Art. 14 | Human oversight | No action executes without verified approval |
| Art. 17 | Quality management | Every receipt is a tamper-evident audit record |
| Art. 26 | Deployer obligations | Receipt log satisfies logging requirements |
A growing body of independent work converges on the same conclusion: AI agents have moved into positions of real authority faster than anyone has built the accountability layer underneath them. AT1C is that layer — and the only open protocol where the human owner, not the platform or the auditor, holds the receipt.
Ten core specification files covering receipts, signatures, scopes, verification, lifecycle, deadman's switch, and staged risk levels. Published open source.
View on GitHub →AT1C uses ML-DSA-65 as its locked signing primitive — the NIST-standardised post-quantum algorithm. Migrated across the full codebase in 2026 ahead of the field.
@at1c/sdk on npm →AT1C is engaged with the W3C Agent Identity Registry Protocol Community Group — the emerging standards body for agent identity. Positioned early, while the group forms.
Position paper →The accountability gap is not a missing technology. As independent researchers have noted, agents are moving into positions of real authority faster than anyone is building the layer underneath them — "the space between a mistake and its consequence, the space a human used to stand in, closed." AT1C reopens that space. Cryptographically. For every person, not just those with enterprise compliance budgets.
AT1C is more than a compliance tool. It is the foundation of a model where humans own their digital identity — across every system, every AI, and every lifetime.
Your verified self — provable, portable, not owned by any platform.
Your anonymous space — protected, untracked, under your control.
Inheritable. Transferable. Your digital life outlives you on your own terms.
Identity belongs to the individual, not the platform. AI agents should serve humanity, not control it. Ownership of self is the foundation. Freedom is the AT1C Protocol. Humans first.
No platform owns it. No institution brokers it. The receipt lives with you.
Register your first agent in minutes. The private key stays on your device. AT1C never holds it.