People, people working with AI, and AI agents now use the same apps. Credia Proof-of-Actor tells you which one is behind an interaction, whether a human is really in control, and who authorised an agent, so your application can decide what to allow.
Available to selected teams. Request access to discuss your integration.
Credia Proof-of-Actor
Status: Private Beta
See if the traffic in your app is human, AI-assisted or agent traffic.
You run a web app, mobile app or API, and you want to know if the traffic in your app is human, AI-assisted or agent traffic. You do not need to build an agent or use a Credia wallet.
Start with Credia Payments to give it a wallet with spend limits, then add Identity, Auth, Verify and Reputation as you need them.
For years, apps could assume a person was on the other side of the screen. That is no longer safe: AI can write the message, answer the test, operate the browser and make the purchase.
A login tells you who can access an account. It does not tell you who, or what, is acting right now. Credia Proof-of-Actor fills that gap: it gathers evidence about each interaction and tells your application whether a human, a human working with AI, or an authorised agent is behind it, with a confidence for each claim.
Sign-up, an online test, a checkout, a payout change: wherever knowing the actor changes your decision.
Request only the checks that moment needs, for example a present human in continuous control, or a valid agent mandate.
You get an Actor Attestation: the actor type and a confidence for each claim, never a hidden yes or no.
Allow, ask for an extra check, send to review or decline. Start by observing before you enforce anything.
Situations application teams face today, and how Credia Proof-of-Actor helps in each.
Candidates can now paste questions into an AI, or let someone else take over the session. A pass no longer means much.
How it works
Hiring teams can rely on assessment results again, and flag only the sessions that need a second look.
A customer logs in, then the session is handed to someone else, remote-controlled by a scammer, or driven by an agent. Changing payout details is where the damage happens.
How it works
Consequential changes need a real, present person, without adding friction to every login.
AI agents are starting to buy on behalf of people and companies. You want the sale, but you need to know who is behind the agent and whether this purchase is allowed.
How it works
Legitimate agents can buy openly instead of pretending to be humans, and you know who is accountable.
Fake accounts flood reviews, polls and free-trial offers. You need one real person per account, not their passport.
How it works
Fewer fake accounts, with no extra personal data collected from genuine users.
Your users write with AI help. Blocking it would hurt them, but some contexts, such as certified work, need to know.
How it works
AI help stays welcome, and the contexts that need transparency get it.
Deepfakes and AI-run profiles make it hard to know who you are really talking to.
How it works
Users get a signal they can trust, based on evidence rather than a profile photo.
You suspect agent and AI-assisted traffic is changing how your app is used, but you do not want to block anyone on a guess.
How it works
Policy decisions based on measured evidence, with no surprise impact on real users.
The same path for a person, a person with AI, or an agent.
It states what this moment requires, such as a present human or a valid agent mandate.
Credia Proof-of-Actor reasons over several independent signals across the session, not a single puzzle.
The actor type and a confidence per claim, bound to this interaction and time.
Allow, step up, review or decline. Credia Proof-of-Actor never makes the decision for you.
A readable Actor Trust Receipt is available for reviewers, audits and the person involved.
Which signals are used, and how they are weighted, is not published. What each result establishes always is.
An example using the Proof of Actor SDK, which carries forward into Credia Proof-of-Actor. API keys, the full reference and onboarding support are provided to beta teams.
import { ProofOfActor } from "@proofofactor/sdk";
// A person: require presence and continuous control
const actor = await ProofOfActor.verify({
require: ["human_presence", "human_control", "session_continuity"],
});
if (actor.humanControl > 0.9) {
allow(); // your policy, your threshold
}
// An agent: is this action inside its mandate?
const authority = await ProofOfActor.verifyAuthority({ agent, action });Credia Proof-of-Actor does not treat every automated interaction as an attack. It makes agency explicit, then lets each application decide what it accepts.
A person appears to be directly controlling the interaction. Example: a candidate completes an assessment personally.
A person stays responsible, but AI materially helps. Example: a professional drafts an email with an AI writing assistant. Often entirely acceptable.
An AI agent acts for an identified person or organisation, within delegated authority. Example: a shopping agent buys a flight within its user's budget.
There is not enough evidence to establish the actor, human control or who is accountable. This is not an automatic finding of abuse.
Credia Proof-of-Actor answers three separate questions. Ask one, or combine them, depending on what the moment needs.
Proof of Human: evidence that a real, present human is behind the interaction, with identity optional and disclosure limited to what you need.
Proof of Control: evidence that the same human is present and controlling the session throughout, and how much AI is participating.
Proof of Authority: which agent is acting, who it represents, and whether this specific action is inside the mandate its principal granted.
Credia Proof-of-Actor reports AI involvement on a scale, so a declared writing assistant is never treated the same as an autonomous agent.
Nothing suggests AI materially shaped the interaction. This is not proof that no AI was used.
Some evidence, not enough to say more.
Evidence suggests AI materially assisted.
The person disclosed their AI use.
An agent, not a human, performed the interaction.
Each result states what was established, not the raw data behind it. It is designed not to carry recordings, documents or biometric templates. "Identity verified, kept private" is a complete answer: your application learns only what its policy requires.
Results are bound to a specific session or action and time, so old evidence cannot be reused somewhere else.
It never claims an interaction is 100% human, and it cannot see every form of AI use: help from another device or a copied answer may not be observable. Unknown results stay visible as unknown. A label does not grant access or block anyone by itself; your application decides. Coverage is confirmed separately for web, mobile and API integrations.
Credia Proof-of-Actor carries forward Proof of Actor within AgentCredia. Proof of Human, Proof of Control and Proof of Authority remain the names of its three checks, and the Actor Attestation and Actor Trust Receipt remain its output. Consult the migration guide before changing existing integrations.
Teams that run web apps, mobile apps, SaaS products, marketplaces and APIs used by people and agents. Product, trust and safety, risk and security teams use the results.
No. A CAPTCHA asks whether a visitor can solve a puzzle once. Credia Proof-of-Actor builds evidence about a real, present human across the session, and gives legitimate, declared agents their own lane instead of forcing them to pretend to be people.
No. You decide. A writing tool may welcome declared AI assistance; a hiring assessment may require a human with low AI assistance. Credia Proof-of-Actor supplies the evidence and your policy makes the call.
No. Off-device help, copied answers and other activity it cannot observe limit what can be established. Results carry a confidence, and unknown stays explicit.
Not by default. Many uses only need "verified human, identity not disclosed". Identity is shared privately, and only when your use case requires it.
No. A result states the claims that were established. It is designed not to carry recordings, documents or biometric templates.
No. Credia Proof-of-Actor is built for the application receiving the interaction and works on its own. You do not need a wallet or a Payments plan.
Credia Proof-of-Actor is in Private Beta. Pricing is agreed with each team during onboarding, based on the checks used and interaction volume. Request access and tell us which journey you want to understand first.
Establish an agent's identity and its connection to an operator.
Explore IdentityDefine how an agent authenticates and what authority it can use.
Explore AuthInspect specific claims and the evidence behind them.
Explore VerifyUnderstand activity and outcomes in context over time.
Explore ReputationPayment workflows and controls for AI agents.
Explore PaymentsTell us which journey matters most: sign-up, an assessment, a checkout or a high-risk action.