A receipt forevery voice.
Every clip ECHO accepts will be issued a public receipt linking the audio to the contributor’s consent, the licence they agreed to, and their payment. Anyone can check one.
Files are hashed on your device. Nothing is uploaded.
Live receipts
Receipts appear here seconds after a clip is accepted.
- 0
- Receipts issued
- 0
- Consent records
- 0
- Payouts linked
- 25
- Controls tracked
Computed from the ledger at build time.
| Event | Receipt | Language | When |
|---|
What a receipt proves
Recorded
The contributor reads a prompt in the app.
Consented
The task’s purpose and licence version are attached.
Hashed
The audio’s SHA-256 becomes its permanent address.
Reviewed
Quality and transcript checks, by machine and by people.
Paid
The contributor’s payout is linked to the receipt.
Licensed
Datasets ship with a manifest of every receipt inside.
{
"receipt": "sha256:9f2c…e41a",
"event": "registration",
"language": "yo-NG",
"task": "scripted",
"speaker": "spk_7Q…", // pseudonymous
"consent": {
"licence": "echo-train-eval-v1",
"agreed_at": "…",
"purpose": "speech recognition training"
},
"review": "accepted",
"payout": "paid"
}Where we stand, stated plainly.
ECHO is pre-launch. Nothing below is claimed as certified until it is, and each line will carry its date and evidence when it changes.
GDPR
In developmentEU General Data Protection Regulation
Consent, access, export and deletion flows are being built into the contributor app.
CCPA / CPRA
In developmentCalifornia Consumer Privacy Act
Same rights model as GDPR, applied to California residents.
BIPA
PlannedIllinois Biometric Information Privacy Act
Voice can be biometric data. Written consent and retention schedules before any Illinois collection.
SOC 2 Type II
PlannedIndependent controls audit
Audit window to begin once production infrastructure is in place.
Independent security review
PlannedThird-party penetration test
Before the contributor app opens to the public.
Public receipt ledger
In developmentEvery accepted clip gets a public receipt
This page. Receipts appear here from the first accepted contribution.
Audit controls
Every control behind a receipt, grouped by what it protects, each with its real status.
Contributor identity
Who can contribute, and how we know they are who they say.
3 in development2 planned
Contributor identity
Who can contribute, and how we know they are who they say.
Identity verification
Government ID matched to the contributor once, at onboarding, before any receipt is issued.
Before first payoutPlanned
Liveness check
A short spoken challenge at the start of each session confirms a live person is recording.
Every sessionPlanned
Own voice only
Contributors attest that every recording is their own voice. Duplicate and synthetic voices are screened before acceptance.
Per clipIn development
Anti-coercion limits
Session and daily limits, device checks and attestations make bulk or coerced recording impractical.
In development
De-identification
Identity never reaches a receipt or a dataset. Each contributor is an opaque speaker ID.
By designIn development
Consent and rights
What a contributor agrees to, and how that agreement is kept.
5 in development
Consent and rights
What a contributor agrees to, and how that agreement is kept.
Per-task consent
Each task states its purpose, licence and pay before recording starts. Consent is captured per task, not once at sign-up.
Per taskIn development
Versioned licences
Every receipt records the exact licence and policy version agreed to. Old versions stay published at their original hash.
Content-addressedIn development
Right to withdraw
In the app, at any time. Unlicensed clips are deleted; licensed clips leave all future dataset deliveries. The receipt records the withdrawal.
ImmediateIn development
Right to access and export
A full machine-readable export of everything held against a contributor's account.
In development
Pay-through
Every accepted clip is paid. Payment status is visible on each receipt.
Per receiptIn development
Data handling
What is collected, what is kept, and what is never shared.
3 live2 in development1 planned
Data handling
What is collected, what is kept, and what is never shared.
Pseudonymous datasets
Datasets carry a speaker ID and self-described attributes only. Never names, contact details or exact location.
In development
Sensitive signals stay private
IP addresses, GPS, device identifiers and audio fingerprints never appear on receipts. Receipts record only whether a check passed.
Existence flags onlyIn development
Minimal website data
Forms collect only what is needed to reply. IP addresses are stored as keyed one-way hashes and purged within 24 hours.
Live
No advertising use
Personal data is never sold or used for advertising.
Live
Retention schedule
Waitlist data is deleted 24 months after last contact, or on request.
Live
Data residency
EU contributors' audio stored in the EU; other regions stored in-region where required.
Planned
Security
How systems and data are protected.
3 live1 in development1 planned
Security
How systems and data are protected.
Encryption in transit
All traffic served over TLS.
TLSLive
Abuse protection
Signed form tokens, honeypot fields and per-connection rate limits on every form.
Live
Encryption at rest
Audio and metadata encrypted at rest in production storage.
In development
Least-privilege access
Production access limited by role, logged and reviewed.
Planned
Vulnerability disclosure
Security reports are accepted through the contact page and acknowledged by a person.
Live
Audit and verification
How anyone can check the claims on this page.
1 live2 in development1 planned
Audit and verification
How anyone can check the claims on this page.
Content-addressed receipts
Every accepted clip is identified by the SHA-256 hash of its audio, so anyone holding the file can verify it here.
SHA-256In development
Public ledger
Receipts are published as they are issued. Withdrawals append an event; nothing is silently removed.
Append-onlyIn development
Dataset manifests
Each delivered dataset ships with a manifest listing the receipt of every clip inside it.
Planned
Live posture
Every number on this page is computed from the ledger at request time, never typed in.
ComputedLive
What ECHO will collect
Every receipt declares what its file is. Identity and location never leave ECHO’s systems.
File specification
- Delivery audio
- WAV, 48 kHz, 24-bit, mono
- Capture
- Lossless on device, uploaded over TLS
- Transcripts
- UTF-8, verbatim and normalised
- Metadata
- JSON, one record per clip
Declared on every receipt
- Language and locale
- Task type and prompt ID
- Duration
- Recording environment
- Device class
- Self-described age band
- Self-described accent
- Licence version
- Review outcome
- Payout status
Never on a receipt
- Name or contact details
- Exact location or GPS
- IP address
- Device identifiers
- Government ID data
- Raw identity documents
Consent and policy versions
Each receipt will name the exact version a contributor agreed to. These are the versions published today, with the hash of their source so any change is detectable.
Security posture and subprocessors: Trust Center
Questions
Something we haven’t covered? Ask us and we’ll add it here.
Is ECHO GDPR compliant today?
ECHO is pre-launch and does not yet collect voice data. The contributor app is being built around GDPR's requirements: explicit per-purpose consent, access, export, deletion and EU data residency. We'll mark it Live here only when it's operating.
Is a voice recording biometric data?
It can be. Voice can identify a person, and laws like Illinois' BIPA treat voiceprints as biometric. ECHO collects written consent per task, never builds voiceprints for identification, and publishes its retention schedule.
How does a contributor withdraw consent?
In the app, at any time. Unlicensed clips are deleted. Licensed clips are removed from every future dataset delivery, and the receipt gains a withdrawal event so buyers can see it.
Why put receipts on a public page?
An audit trail only the seller can see isn't an audit trail. A public ledger lets contributors, customers and regulators check the same record on the same terms.
What's on a receipt, and what isn't?
The audio's hash, language, task, licence version, review outcome and payout status. Never identity, location, IP address or device identifiers. Those stay in ECHO's systems, and receipts carry only a pass/fail flag.
Can I see the recordings themselves?
Not through this page. Receipts prove provenance without exposing audio. Teams evaluating a dataset get access through a commercial agreement.