R1
R
ecruiter
1

Math,
not messages.

The server stores math, not messages. Every conversation is end-to-end encrypted with AES-256-GCM. Keys are exchanged via X25519. Device identity is bound by TOFU fingerprinting. Every key change is recorded in a public, tamper-evident transparency log.

See how it works
01 · Encryption

Protected by construction.

Three properties that make R1 secure without trusting the server.

01

Math, not messages.

The server stores ciphertext only. Every conversation is end-to-end encrypted with AES-256-GCM. Keys are exchanged via X25519. Device identity is bound by TOFU fingerprinting.

02

Keys, client-side.

Device keys are generated client-side and never transmitted in plaintext. Portal access is granted via signed capability tokens, not accounts.

03

Changes, logged.

Every key change is recorded in a public, tamper-evident transparency log. Checkpoints are externally anchored to an independent timestamping service.

02 · Live demo

Real cryptography.

Not a faked scramble. Real in-browser WebCrypto — X25519 key exchange, AES-256-GCM encryption.

AC
Acme · Senior Engineer
Direct message
On LinkedIn
Great call today — the team loved you.
Likewise. Excited about it.
What's the real ceiling on base?
“Let's move this to Signal.”
Stored readable · AI-scanned · subpoena-discoverable.

Every other network makes you leave. R1 doesn't.

03 · Infrastructure

Hardened at the edge.

Every response includes HSTS, strict CSP, X-Frame-Options, X-Content-Type-Options, and CORP/COOP/COEP isolation.

Independent Verification

Our production environment is publicly audited. We maintain a top-tier rating for network protocol resilience and data-injection mitigation. In production, every response includes HSTS, strict CSP, X-Frame-Options, X-Content-Type-Options, and CORP/COOP/COEP isolation.

A+Security Ratingsecurityheaders.com

  • max-age=31536000; includeSubDomains; preload

    HSTS Enforced.

    All browsers must interact with R1 over encrypted HTTPS. Downgrade attacks and cookie-hijacking are blocked at the protocol level.

  • script-src 'self' + nonce-based

    Strict CSP Active.

    Content-Security-Policy restricts script execution to trusted sources only. XSS and unauthorized data injection vectors are eliminated.

  • frame-ancestors 'none'

    Frame-Ancestors Locked.

    R1 cannot be embedded in iframes by anyone. Clickjacking and UI redressing attacks are blocked via both CSP frame-ancestors and X-Frame-Options headers.

  • worker-src 'self'

    Worker-Source Restricted.

    Service workers and Web Workers can only originate from R1's own domain. Third-party code execution in worker contexts is blocked.

    • Verify Our Live Security Profile

    Don't take our word for it. Run our domain through the independent Security Headers framework and see the full report.

    Run Independent Audit
    04 · Threat model

    What's protected.

    R1 protects message content, candidate profiles, and decision reasoning from server-side access. The server cannot read encrypted DMs, portal chat messages, or candidate communications — it stores ciphertext only.

    Protected
    Signals DMs
    Portal chat
    Team DMs
    Candidate profile fields
    Decision reasoning
    Not protected
    Job board listings (public by design)
    Aggregate job counts
    Sitemap data
    Newsroom content
    Server-visible metadata
    Message timestamps
    Participant counts
    Applies everywhere.

    These protections apply across all R1 products — Jobs, Signals, Portals, and Application Studio. For implementation details, see the sections below.

    05 · The broken default

    The old signals are gone.

    Résumés, interviews, profiles — all generated. All polished. All worthless.

  • 01 · Access Control

    Your data becomes training data.

    CVs become PDFs. PDFs become training data. Once it's sent, it's gone — and AI has it forever. 0% access control.

  • 02 · Avg Time-to-Hire (Interview Guys '25)

    Portals kill momentum.

    Account. Dashboard. Training. By the time they're in, the candidate is gone. 44 days average time-to-hire.

  • 03 · Version for All

    Every profile looks the same.

    AI can replicate any profile at scale. Nothing is verifiable. Nothing is device-bound. The same polished CV goes everywhere. 1 version for all.

  • 04 · Update Loop

    Files die the moment they're sent.

    A PDF is frozen the moment it's downloaded. No updates, no context, no way back to the source. No update loop.

    • 06 · The R1 answer

    Connect Protocol.

    A single URL replaces the entire PDF-email-login-portal pipeline. Here's what it does.

    Acme CorpSenior Engineer

    Alex Chen

    Senior Engineer

    Led the real-time sync engine rewrite at Linear, reducing p99 latency by 68%. Architected distributed event sourcing pipeline processing 2.4M events/day with zero data loss across 3 regions.

    Experience

    Staff Engineer — Real-time Systems
    Linear
    2022 — Present
    • Rewrote the sync engine from polling to CRDT-based real-time, cutting p99 from 1.2s to 380ms
    • Designed event sourcing pipeline handling 2.4M events/day across US, EU, APAC regions
    ReactNode.jsDistributed SystemsCRDTsEvent SourcingAWSPostgreSQLKubernetes
    Senior Engineer — Platform
    Vercel
    2019 — 2022
    • Built the edge function runtime serving 14B requests/month with sub-50ms cold starts

    Skills

    ReactNode.jsDistributed SystemsCRDTsEvent SourcingAWSPostgreSQLKubernetes
    Acme CorpAlex Chen
    Senior Engineer
    Lisa Park

    Alex looks strong. What's their availability?

    2m ago

    Available in 2 weeks. Also interviewing at Stripe — we should move fast.

    1m ago
    Lisa Park

    Let's schedule for tomorrow. Can you confirm?

    30s ago

    Done. Calendar invite sent. I'll prep Alex tonight.

    Just now
    Type a message...
    E2EE
    R1Connect
    07 · Honest comparison

    R1 Connect.

    A live candidate page, backed by the R1 talent network.

    Capability
    R1
    Connect
    PDF
    Email
    ATS
    Portal
    LinkedIn
    Profile
    Opens without an account
    		VariesVaries
    Tailored to the role
    		ManualVaries—
    Expiring shared access
    		—Varies—
    Viewer limit for a shared link
    		—Varies—
    Trust phrase for a new browser
    		———
    Chat embedded on the candidate page
    		—Varies—
    End-to-end encrypted embedded chat
    		—Not typical—
    Key numbers highlighted automatically
    		ManualVaries—
    PDF linking back to the live page
    		ManualVaries—
    Platform cannot read message content
    		———
    Offline copy
    		Varies—
    Compare and rank candidates for a role
    		ManualVaries
    Built-in candidate network
    		——
    Network size
    		Growing——1B+
    Built in
    —Not offered
    R1 exclusive
    08 · Message Security

    Private by default.

    Recruiting platforms hold your most sensitive professional data. In 2025, incidents put 100M+ records at risk.

    Sources: McHire researcher writeup, TalentHook / Cybernews, Foh&Boh / Cybernews, HireClick

    100M+
    Records put at risk
    Recruiting platforms, 2025
    $10.22M
    Average US breach cost
    Source: IBM Cost of a Data Breach 2025, CyberScoop coverage
    €1.2B
    GDPR fines in Europe
    Source: DLA Piper GDPR survey 2026
    0
    Plaintext messages we can read
    Source: R1 client-side encryption + server-side validation
    See it happen
    Your device
    Plaintext, on your screen
    AES-256-GCM · X25519
    R1 servers
    —
    Everything R1 can see
    Recipient device
    Decrypted, for their eyes only
    Awaiting trusted device

    Encrypted on your device. R1 stores ciphertext. Only the recipient's device can read it.

    What keeps happening

    The breaches so far.

    Real incidents. Real numbers. Real exposure.

  • 01 · McDonald's franchisee hiring platform

    McHire — 64M records.

    A hiring chatbot used by McDonald's franchisees exposed a path to applicant data through basic access-control failures, including default credentials and an IDOR vulnerability. Researchers reported that up to 64 million applicant records could have been accessed. Source: Ian Carroll + Sam Curry, WIRED.

  • 02 · Applicant Tracking System Leak

    TalentHook — 26M files.

    A misconfigured Azure Blob container exposed nearly 26 million files, mostly resumes and CVs, with personal and professional details. Source: Cybernews.

  • 03 · Hospitality and retail hiring platform

    Foh&Boh — 5.4M files.

    A publicly exposed AWS bucket contained 5.4 million files, largely resumes and CVs submitted through a hiring platform used by hospitality and retail brands. Source: Cybernews.

  • 04 · Resume platform breach

    HireClick — 5.7M resumes.

    A misconfigured cloud bucket reportedly exposed 5.7 million resume files containing names, addresses, emails, phone numbers, and employment history. Source: Daily Security Review.

    • Encryption comparison

    What protects the message.

    Each row is written in plain English: what it means for you, not what it means for a cryptographer.

    Protection
    R1
    Protected chat
    LinkedIn
    Messages
    ATS
    Messaging
    WhatsApp
    Personal
    Encrypted in transit with TLS
    Your messages are scrambled while traveling across the internet.
    Messages encrypted end to end
    Only you and your recipient can read the message — not even R1.
    		——
    Service cannot read message content
    We can't see your conversations, even if subpoenaed.
    		——
    Keys unique to each device
    Each phone or laptop gets its own lock — one compromised device doesn't expose the rest.
    		——
    Thread keys wrapped for each recipient device
    Adding a new phone doesn't expose old messages to it.
    		——
    Trusted key changes are detected
    You're alerted if someone's device keys change unexpectedly — a sign of tampering.
    		——
    Append-only audit log of every device key change
    Every key change is recorded permanently — you can verify the full history yourself.
    		——
    Revoked devices stop receiving message keys
    Lost a phone? Remove it and it instantly loses access to new messages.
    		——
    Cross-device recovery without exposing plaintext to the service
    Add a new device without us ever seeing your message content.
    		——
    Unverified cross-party devices blocked before key delivery
    Unknown devices from other parties are blocked before they can receive your keys.
    		———
    Invalid encrypted envelopes rejected before storage
    Tampered or malformed messages are rejected before they're ever stored.
    		———
    Realtime message alerts contain metadata only
    Push notifications say 'New message' — never the actual content.
    		———
    Built in
    —Not offered
    R1 exclusive

    ATS refers to standard native messaging in Greenhouse, Lever, Bullhorn, iCIMS, and Workday. Third-party integrations are excluded.

    Protections apply to message content. Metadata — who, when — is not message content. Every key change is logged and client-verifiable.

    09 · How it works

    The engine underneath.

    The few things that keep your data yours.

  • Single-use link with configurable expiry

    One link. One key.

    Every recruiter link is a one-time credential for one candidate and one role. When it dies, the access dies with it.

  • 15-word device-bound connect phrase

    The right device gets the phrase.

    First viewer receives a device-bound connect phrase. New devices need it. Too many wrong tries lock the door.

  • Viewer limit per shared link

    You control the circle.

    Set a limit on how many people can hold the link at once. Once it's full, no one else gets in.

  • Time-bound link expiry

    It expires when you say.

    Configurable countdown. Clean 'not found' after expiry. No stale data, no ghost access, no surprises.

  • One-time magic token in PDF

    The PDF stays alive.

    Every PDF download carries a one-time token back to the live, controlled page. The offline file never becomes a dead end.

  • Embedded E2EE chat

    Real talk.

    Hiring managers chat right on the candidate page. No Slack. No email. No context lost.

  • Client-side AES-256-GCM + HKDF

    Encrypted before it leaves.

    Your message is locked on your device. The server only ever handles noise. Sources: R1, NIST GCM, RFC 5869 HKDF.

  • X25519 key agreement per device

    One key per device.

    No shared master key. No single breach that unlocks everything. Each device gets its own sealed copy. Sources: R1, RFC 7748 (X25519).

  • Metadata-only realtime events

    Realtime only knows activity.

    It sees that something happened. Never what was said. Never the body.

  • Device-scoped identity keys

    Every device has its own identity.

    Private keys stay local. The server only knows public keys. Revoke one device without signing out everywhere. Sources: R1, RFC 7748, RFC 8032.

  • Verified-recipient enforcement

    Keys go to verified devices only.

    Device, thread, trust status — all checked before any key is stored or delivered. No exceptions.

  • TOFU fingerprinting with SHA-256

    First trust, forever.

    We pin the first fingerprint of every device. If the key changes, messaging stops until you verify.

    • The Bottom Line

    AI broke hiring. R1 rebuilds it on trust.

    Message content is encrypted on your device. The server stores math. It cannot read it, mine it, or train on it. Your offers, your counteroffers, your rejections — yours.

    • One device, one key.
    • Encrypted before it leaves.
    • Verified before it's delivered.
    • Pinned on first trust.
    • Revoked in one click.

    Messages reach the recruiter and no one in between. Not even us. Not a privacy policy. Math. See the transparency log →

    Hire(d) better.

    Security — R1 Trust Infrastructure