Open questions
What we are working on
These questions shape the infrastructure. If you research any of them, we would like to hear from you.
Measuring human control without surveillance
What is the minimum evidence that supports a confident claim of human control, and how should confidence be communicated so it is never mistaken for certainty?
Graded AI participation
How should AI involvement be expressed across a spectrum from none to autonomous, in a way that is useful to policy and fair to people who use AI openly?
Portable, revocable delegation for agents
How can a principal grant an agent scoped authority that third parties can verify per action, and withdraw it instantly?
Selective disclosure in attestations
How can an attestation prove "verified human" or "authorised to spend up to $500" without revealing who the human or principal is?
Continuity across a session
How should a trust claim decay or persist as a session continues, so evidence gathered at login is not mistaken for evidence about later actions?
Interoperability with emerging standards
How should actor attestations relate to existing identity, credential and agent-protocol standards rather than competing with them?
Essays
Foundational essays
Published in English.
Differentiation
Adjacent categories, and the question they leave open
Each of these is valuable. None alone answers who or what is actually acting, how much human control exists, and who is accountable.
| Approach | Answers | Proof of Actor also asks |
|---|---|---|
| Identity verification | Who are you? | Who is acting right now? |
| Bot detection | Does this traffic look automated? | Is automation occurring, is it legitimate, and who stands behind it? |
| Proof of personhood | Are you a real human? | Are you currently present and controlling this interaction? |
| Agent identity | Which agent is this? | Who authorised it, what authority does it have, and is this action inside that authority? |
| Deepfake detection | Does this media appear synthetic? | Regardless of how the media was produced, who is actually responsible for this interaction? |
| Authentication | Can this entity access this account? | Who or what is actually performing this action right now? |
An emerging, not empty, category
Market activity validates the need. Investors and builders increasingly describe proving you are human online as a major internet trust problem, and existing approaches cover adjacent pieces:
- Identity and authority infrastructure for AI agents
- Consumer-to-agent authorisation services from credit and identity providers
- Know Your Agent and agent-to-human binding from verification providers
- Verified bots and signed agents at the network edge
- Continuous human verification
- Proof of personhood systems
Our focus is the broader Actor problem: not merely who owns an identity and not merely whether automation exists, but who or what is actually acting in a specific interaction, how much human control exists, and who is accountable.
Standards we build alongside
Proof of Actor is designed to complement existing and emerging work, not replace it.
- WebAuthn and passkeys
- Phishing-resistant authentication that proves possession of a credential.
- W3C Verifiable Credentials
- A data model for cryptographically verifiable claims and selective disclosure.
- OAuth 2.0 delegation
- Established patterns for granting scoped access to applications.
- Remote attestation (IETF RATS)
- An architecture for conveying evidence about a system's state to a relying party.
- HTTP message signatures for bots and agents
- Emerging work letting automated clients identify their operator cryptographically.
- Agent payment and commerce protocols
- Industry protocols for how agents hold credentials and complete checkout.
Researchers and standards communities
Identity, WebAuthn, agent protocols, privacy, cryptography and trust infrastructure. Follow the research or start a conversation.
Follow the Research