The verifiable proof layer for regulated enterprise AI

A log is a claim. A receipt is proof.

AskLedger turns every AI decision into a signed receipt, so when a regulator, auditor, or court asks what your AI did, you can prove it independently, without asking anyone to trust you.

Built for banking, insurance, and healthcare, and for the agentic AI now acting on its own.

$ npx @askledger/receipts-sdk verify receipt.json
signed-receipt.jsonSIGNED · VERIFIED ✓
receipt_id01J9X8VKHT…
tenant_idacme-bank
event_typegateway.request
ai_modelclaude-sonnet
chain_height12,487
prev_hash92cd1a0b…d04e5e
receipt_hasha73f2ce5…8b3c1e
timestampRFC 3161 · DigiCert TSA
signatureEd25519 · valid
The accountability gap

Enterprises are deploying AI faster than they can prove what it did.

AI is now in nearly every business function, incidents are climbing, and transparency is falling, exactly as regulators begin to demand evidence of what each system decided and why. The obstacle is rarely the model. It is that organizations have no reliable, independent way to prove what their AI actually did when an auditor, regulator, or court asks.

FigureWhat it meansSource
88%of organizations now use AI in at least one business functionStanford HAI, 2026
$581.7Bin global corporate AI investment in 2025, more than double the year beforeStanford HAI, 2026
362AI-related incidents logged in 2025, up 55% year over yearStanford HAI, 2026
40Foundation Model Transparency Index score out of 100, down from 58 the year beforeStanford HAI, 2026

AskLedger closes that gap. It turns each AI decision into a signed, tamper-evident receipt, so you can prove what your AI did in evidence anyone can check, without asking them to trust you. See how → And what it does not prove →

The problem

Five questions about your own AI.
Most companies can't answer a single one.

From a solo builder to a global enterprise, AI is now everywhere in the work, and almost no one can say what it's actually doing. Ask yourself:

01

Who, or what, is using AI?

Your people, your apps, service accounts, and agents acting on their own.

02

Which tools, models and data?

Every model, data source, and permission actually in play.

03

What is it really costing?

The true bill: subscriptions, usage, retries, review and rework.

04

What risk or result did it create?

The business outcome, or the exposure, behind each AI action.

05

Can you prove any of it later?

When an auditor, regulator or customer asks, you need proof, not a claim.

AskLedger answers the one question that makes the other four matter: can you prove it? Without cryptographic proof, the other four are just claims. With it, they are evidence anyone can check.

Why it's different

Not another logging or governance tool.

A signed receipt for every AI decision, across five language SDKs, hash-chained and independently verifiable. Not a log you keep and could edit, but portable evidence an outsider can check without trusting you.

Independent

Verify it yourself, trust no one

Verifiable by anyone with a public key. No account, no trust in us, no trust in the vendor. Change one character and verification fails.

This is the whole idea: evidence, not a claim.
Open

A trust layer you can inspect

Open source, Apache-2.0, built on published RFCs. A trust layer you have to take on faith is not a trust layer, so ours is inspectable, end to end.

Vendor-neutral

Works with your stack, not instead of it

Any model, any provider, any gateway, any region. It sits above what you already run instead of replacing it.

Confidential

Proof without exposing prompts

Receipts record a hash of the input and output, not the raw content. The private signing key stays inside your own HSM.

Proof is the core. Around it sits a five-layer accountability model, from preventing the wrong action to enabling the program that governs it. See the five layers →

Integrate, don't replace

Keep your gateway. Keep your observability. Keep your security stack.

AskLedger sits above the tools you already run, importing their traces, costs and signals, then adding identity, policy, outcomes and independent proof. We connect your stack; we don't ask you to rip it out.

PortkeyLangfuseHeliconeOpenTelemetryOpenAIAnthropicLangChainn8nOktaMicrosoft EntraAWS · Azure · GCP bills
How it works

The whole idea, in three pictures.

An AI decides something

Like declining your loan or flagging a payment. It happens in a split second.

A receipt is made automatically

It records exactly what the AI did, and when. Just like the receipt you get at a shop.

Anyone can check it later

A regulator, your bank, or you. Change even one word and it stops verifying.

That's AskLedger: a receipt for what AI does.

Don't believe it? Fake it yourself.

Pick an example, watch AskLedger lock it, then try to change it and see what happens. Real cryptography, running right on your device.

1Your AI makes a decision
2AskLedger turns it into a locked receipt
Try the open tools

Sign and verify a receipt yourself. No signup.

The SDK, the CLI and the verifier are open source under Apache-2.0. Three ways to try them right now:

In your browser

Playground

Sign a receipt, tamper with it, and watch verification fail, live in your browser. Nothing to install.

Open the playground →
In your terminal

The CLI

Verify any receipt from the command line, no install needed.

$ npx @askledger/receipts-sdk verify receipt.json
CLI reference →
In your code

The SDK

One import, five languages, an independent verifier in every SDK.

$ npm install @askledger/receipts-sdk
For developers →
5
Language SDKs, wire-compatible
<1 ms
Cryptographic signing cost, off the request path
63/66
Hardening controls verified, 2 unbuilt, 1 open
0
Network calls to verify
Works with OpenAIAnthropicGeminiBedrockAzureCohereMistralHugging FaceGroqLangChainCursor
Why now

The clock has already started.

India's RBI has closed consultation on its Model Risk Management guidance, the UAE CBUAE deadline follows within weeks, and the EU AI Act's high-risk logging duty lands in December 2027. Each one asks for evidence of what an AI system did. It's the record most institutions can't yet produce.

India · RBI, Model Risk Management
Public consultation on the draft Model Risk Management guidance closed 24 July 2026. The final circular is expected, and binds banks and NBFCs once notified.
Act now · banks & NBFCs in scope
European Union · AI Act
High-risk obligations (Article 12 logging & Annex IV), from 2 December 2027
Days
Hrs
Min
Sec
UAE · CBUAE Federal Decree-Law No. 6
Transitional deadline for regulated financial institutions
Days
Hrs
Min
Sec
Why AskLedger

If your company runs AI, you need proof of what it did.

It isn't only regulated industries. Any team putting AI into real decisions gets the same advantages, starting on day one.

Win enterprise deals

Pass security reviews and RFPs with cryptographic proof a buyer's own team can check, instead of a claim they have to take on faith.

Ship AI faster

Accountability becomes one line of code, not a six-month compliance project. Stop stalling in pilots.

Make AI risk insurable

Verifiable evidence lets you file a claim, price a policy, or hold a vendor liable when AI gets it wrong.

Control third-party AI

When a vendor's model touches your data, prove exactly what it did, and keep accountability where it belongs.

Debug & monitor agents

As agents act on their own, get a tamper-evident record of every action for forensics, drift, and control.

Future-proof compliance

One evidence layer supports the record-keeping obligations across the frameworks coming: EU AI Act, RBI, NIST AI RMF, ISO 42001. Evidence, not a compliance guarantee.

Industries

See exactly how it works for your sector.

Real scenarios, end to end, the decision, the moment it's challenged, and the evidence a receipt gives you.

Regulatory coverage

Mapped to the frameworks your regulators enforce.

EU AI Act, India's RBI, UAE CBUAE, US NIST, UK, APAC, ISO 42001. AskLedger produces the portable, signed evidence these frameworks ask for, built on published RFCs so a receipt holds up across jurisdictions. The mappings are ours, written from each regulator's published text and not reviewed by any regulator. It supports compliance; it does not replace it, and legal sufficiency always depends on your use case and your counsel.

See regulatory coverage → Get the 2026 tracker

"The next five years of AI are not about better models. They are about the infrastructure that makes them defensible."

Built in the open, because trust infrastructure should be inspectable. Open-source core under Apache-2.0; a commercial layer (hosted verification, a regulator portal, and evidence packs) is in design-partner preview.

Questions

What teams ask first.

No. Receipts record a hash of the input and output, not the raw content. The private signing key stays inside your HSM, and verification needs only the public key. Nothing confidential is exposed to produce or verify a receipt.

The cryptographic cost of a receipt is under a millisecond: canonicalize, hash and Ed25519 sign total about 0.4 ms. Receipts are emitted off the request path, so that cost never blocks the AI response, and it is noise against a model call that takes hundreds of milliseconds. Durable persistence is separate: the file-backed reference SDK fsyncs each write for crash safety, which is hardware-dependent and around 10 ms on a laptop SSD, while production uses Postgres where it is a batched commit. Verifying takes about 1.8 ms and normally happens later and offline, when someone audits the record. Reproduce the crypto numbers with npm run bench. Receipt generation never blocks or breaks the AI call it instruments.

No. AskLedger is open source under Apache-2.0, the wire format is an open specification, and any third party can verify a receipt with published RFCs and a public key. The protocol is a public good; adoption is the moat, not lock-in.

OpenAI, Anthropic, Gemini, Bedrock, Cohere, Mistral and more via drop-in adapters, plus LangChain, Cursor, and gateway integrations, across five language SDKs. A generic adapter covers any HTTP-based provider.

As a library inside your own environment, no data routed to a third party. Signing uses your existing AWS KMS, Azure Key Vault, GCP KMS, or PKCS#11 HSM. Chain state runs on your Postgres for multi-tenant scale.

You can build a log. But a record you generate yourself, you can also edit, delete, or backdate, so it won't hold up when someone disputes it. AskLedger produces a signed record that cannot be altered without the change being detectable, and that an outsider can verify without trusting you. Capturing what happened is the easy part; proving it is what we solve.

Ordinary logs are editable and only mean anything inside your own systems. A receipt is tamper-evident (change one character and verification fails) and independently checkable by anyone with a public key. A log is a claim; a receipt is evidence.

Our own format, and we would rather say so plainly. It is an open specification, published and Apache-2.0 licensed, written as a candidate for standardisation. No standards body has adopted it. What is standardised is everything underneath it: RFC 8785 for canonical JSON, Ed25519 for signatures, RFC 3161 for timestamps. So you are not trusting an AskLedger invention to verify a receipt, you are trusting published cryptography plus a spec you can read. We also publish mappings to ISO/IEC 42001, the EU AI Act, India's RBI guidance and NIST AI RMF, which we wrote ourselves from the regulators' published texts; no regulator has reviewed or endorsed them.

Insights

Perspectives on trustworthy AI.

Where AI accountability meets the real world, the decisions, risks, and regulations shaping how enterprises prove what their AI does.

Free resource

The 2026 AI Compliance Deadline Tracker.

Every AI-accountability deadline that matters, EU AI Act, India RBI, UAE CBUAE, NIST AI RMF, ISO 42001, in one place, with the evidence each one demands. Know what's coming before your auditors ask.

  • Every framework, one timeline, dates, scope, and what triggers each obligation.
  • The evidence checklist, exactly what record each regulator expects you to produce.
  • No spam, the tracker, plus the occasional note when a deadline moves.
Partners

Bring AskLedger to your clients.

Product leaders and technical & business consultants already sit where AI-trust decisions get made. Partner with us, help your clients prove their AI, and own the category before it's crowded.

Refer

Send us a lead. When it becomes a customer, you earn, and your client gets the AI accountability layer they're already being asked for. Zero lift.

Implement

Deploy AskLedger for your clients as a new service line. Get certified, own the integration work, and become the go-to expert in your vertical.

Advise

Shape the roadmap as a design partner, co-author case studies, get early access, and build your reputation as an AI-governance authority.

Founding Partners get a directory listing, a partner badge, early access, and co-branded case studies, a limited first cohort while we build the program.

Become a founding partner

Tell us a little about you. We review every application and reply personally, usually within two business days.

Prefer to talk first? Book a call or reach us on GitHub Discussions.

Join the pilot cohort.

We are opening a pilot program for banks, insurers, and AI teams preparing for the EU AI Act, RBI, and CBUAE deadlines. If that is you, let's prove your AI together.

Book a demo