Trust Center

What we hold, and what we refuse to.

This page documents how ClearCiviQ protects your data: the architecture, how your vault is encrypted, the complete list of what we store, and the things we have deliberately built ourselves out of being able to see.

Last updated: November 12, 2026

Architecture

Privacy by design,
not by policy.

PILLAR 01

Zero-knowledge vault

Everything in your vault is encrypted on your device, or in your browser on the work site, before it leaves. Your passphrase never travels to us. The key derived from it never travels to us. Our servers hold ciphertext that is mathematically unreadable without a key that only exists on your devices. This is not a setting you enable. It is the only way the product works.

PILLAR 02

Data minimization

The default for every field is: do not store it. For a personal account we keep a one-way hashed sign-in identifier, your subscription tier, account timestamps, the device type and IP address each sign-in came from and, if you monitor a public license such as a professional license, its number. No email. No name. No password database. A work account also keeps the work email its owner signs in with, their role, and the organization’s activity log. A breach of our stored data would yield encrypted blobs, hashes and those account records, not the contents of anyone’s vault.

PILLAR 03

On-device PII scrubbing

Before your goal text reaches our AI provider, identifiers like names, addresses, and SSNs are redacted on your device or in your browser, then checked again server-side.

PILLAR 04

No training on your data

Your inputs are used to generate your roadmap and nothing else. We do not use your data to train AI models, and our agreements with providers prohibit it.

PILLAR 05

Crypto-shredding deletion

Deleting your account destroys the encrypted blobs tied to it, and without the key that is irreversible by anyone, including us. Deletion is immediate when you confirm it with your passphrase. If you have lost both your passphrase and your recovery key you can still schedule it, and it completes after a 30-day waiting period that any successful unlock cancels. A work account belongs to its organization: it ends when the organization removes you, and an organization that closes is deleted 30 days after it ends.

Cryptography

How your vault is protected.

No proprietary crypto. Every primitive we use is an open, standard, widely reviewed one, from the libsodium library, and we name each of them below.

Key derivation
Argon2id, on your device. Your passphrase is turned into a key with Argon2id at 64 MiB of memory and 3 passes, the OWASP-recommended settings. It is deliberately slow and memory-hungry, which is what makes guessing it at scale impractical. The passphrase itself is never transmitted and never stored.
Vault encryption
XChaCha20-Poly1305. Authenticated encryption with a 256-bit key: every record is sealed with integrity protection, so tampering is detectable rather than merely unreadable. Encryption happens on your device, before anything is sent.
Key separation
Independent keys per purpose. Separate keys are derived from your passphrase key for encrypting your vault and for checking your passphrase on your device. You sign in with Apple, Google or your company’s sign-in, which has nothing to do with your passphrase, so signing in can never derive your vault key.
Recovery
A recovery key generated on your device during setup and held only by you. It can restore vault access if you forget your passphrase. It is never transmitted to us; we store only a copy of your vault key sealed under it, which we cannot open. We cannot reset, recover, or regenerate it.
Sharing
Sealed to the other person’s key. Emergency access, account transfer and shared folders seal a key to the recipient’s public key (X25519 sealed boxes), so only their device can open it.
In transit
Encrypted on every connection between your device and our infrastructure, using current TLS. There is no unencrypted path.
At rest
Two layers. The ciphertext your device produced is stored on Microsoft Azure, which encrypts its disks and storage again with keys Microsoft manages.

We publish the algorithms because a sound design does not depend on keeping them secret. What protects your vault is a key we never hold. One honest limit: to build a roadmap, our servers work with your scrubbed goal, your answers and the finished roadmap in readable form, in memory only, until your device collects it (at most 4 hours). That is the part of ClearCiviQ that is not zero-knowledge, and the inventory below says so.

Data Inventory

What we hold. What we never will.

The complete list of what we store
  • A hashed sign-in identifier. One-way, salted. We cannot reverse it into your identity.
  • Subscription tier. Free, Pro, or Pro Family, synced from your app store, or a personal account included with your job.
  • Account timestamps. Created and last-seen dates, used for billing integrity.
  • Sign-in sessions. The device type and IP address each sign-in came from, so you can spot a sign-in you do not recognize. Deleted 90 days after the session ends.
  • A push token. If you allow notifications, so alerts can reach your phone.
  • Encrypted vault blobs. Your roadmaps and documents as ciphertext we cannot open, with each document’s size and file type.
  • For work accounts. Your work email, your role, your organization’s activity log (who did what, never what is inside a roadmap or document), its plan and invoices, and the company’s licenses, which are public records and are stored readable so we can check them with the issuing agency.
  • Community tips you choose to share. Public, and never linked to your account.
  • Public licenses you monitor. If you ask us to watch a public license, such as a professional or business license, we store its state, type and number, readable and linked to your account, because our servers look it up at the issuing agency. Personal case and document numbers, such as a green card or USCIS case, are never stored.
  • Roadmap quality counts. Anonymous weekly counts of problems found in roadmaps, by kind of process and state, never linked to you. Counts below five are never shown and are deleted after five weeks.
What never touches our servers
  • Your name or personal email. The optional display name in Settings stays on your phone and is never sent to us. Sign-in is handled by Apple or Google: Apple sends no email, and Google’s token includes it, but we read only the account identifier and discard the rest. A work account is the exception: we keep the work email you sign in with.
  • Your passphrase or keys. Generated and held on-device only.
  • Readable vault contents. Documents, notes, and roadmap details are stored in plaintext only on your devices. While a roadmap is being built, and until your device collects it, it exists only in our servers’ memory, never on disk, and is wiped as soon as your device collects it (at most 4 hours). An encrypted backup copy, so a restart does not lose it, is deleted on the same schedule.
  • Biometrics. Face ID and fingerprint stay inside your device's secure enclave. The app receives a pass/fail signal, and nothing about it is sent to us.
  • Advertising identifiers. No ad SDKs, no trackers, no cross-app identifiers, ever.
Operations

How we run the service.

Least-privilege access

Our running services use managed identities with narrowly scoped permissions, and changes reach production only through our deployment pipeline. Production secrets, such as the database, AI provider and signing keys, live in a hardware-backed vault that is closed to the public internet; a few others are stored as encrypted secrets in the hosting platform. None are in code or configuration files. Access to production is limited to the fewest people who need it, each with their own named account, changes to our cloud setup are recorded in its activity log, and every sign-in to our database and every change to its permissions or structure is logged.

Continuous monitoring

Infrastructure health, anomalous traffic, and abuse patterns are monitored around the clock. Monitoring covers system behavior, never the contents of your encrypted data, which we cannot read.

Automated security review on every release

Every deploy is blocked unless static analysis and dependency and code scanning pass, and every production release of the service is preceded by an authenticated dynamic scan of the running service. Findings are remediated on a severity-based timeline, and accepted risks are recorded with an expiry date rather than waived silently. We do not currently publish a third-party penetration test; when we commission one we will say so here rather than imply it.

Vendor minimization

We work with a deliberately small set of service providers, and we name them. Apple provides sign-in, billing and app-integrity checks through the App Store on iOS; Google provides the same through Google Play on Android. Work accounts sign in through the company’s own choice of Microsoft, Google or another identity provider. RevenueCat records which subscription plan an account holds; it receives an opaque account identifier and subscription status, never your name or vault contents. Anthropic, our AI provider, receives only PII-scrubbed goal text, the text of community tips you post (to moderate and tag them), a record number you asked us to monitor with your explicit consent, and a company’s license numbers when it asks us to check them with the issuing agency. That processing runs only in the United States. Microsoft Azure runs our servers and stores ciphertext it cannot read. Cloudflare hosts our website and the business sites and runs our DNS; it sees ordinary web traffic, never vault contents. Google Workspace sends our email, such as sign-in codes and invites; it receives the address and the message. Stripe bills business plans; it receives the business’s name, billing email and usage counts, and customers pay by bank, so no card details. Expo delivers app updates and brokers push notifications; it receives a push token and your device platform, never your name or vault contents. No analytics brokers. No data resellers. This is the current and complete list.

Breach notification

In the event of a security incident affecting your data, we will notify affected users and regulators as required by applicable state law. Because of the zero-knowledge design, a breach of our stored data could not expose vault contents. It could expose what the inventory above says we keep readable: hashes, account and session records, work email addresses, organization records and monitored license numbers.

Compliance

Privacy law, taken literally.

ClearCiviQ is built to comply with the full body of U.S. state privacy law, including California's CCPA/CPRA, Illinois BIPA, Virginia CDPA, Colorado CPA, Texas TDPSA, and the comprehensive laws of more than a dozen other states. Our architecture makes much of this straightforward: rights like deletion and access are easy to honor when the data held is minimal and the rest is encrypted beyond our reach.

Responsible Disclosure

Found a vulnerability?

We welcome good-faith security research. Report findings privately and we commit to acknowledging within 48 hours, keeping you informed through remediation, and never pursuing legal action against good-faith research.

security@clearciviq.app

Please include reproduction steps. Do not access other users' data or degrade the service.