Security, honestly stated.
Security-first design. Best-practice controls. Industry-standard encryption. No procurement-team theater. No certification badges in the footer that we don't have. Here's what we actually do, what we don't, and what's on the roadmap.
What's in place today.
TLS 1.3 everywhere. HSTS preloaded.
Every byte that leaves your browser is over TLS 1.3 with modern cipher suites. HSTS is preloaded. Mixed content is blocked at the CDN edge. We rotate certificates every 60 days.
AES-256, KMS-managed.
Postgres data is encrypted at rest with AES-256 on our managed Postgres sub-processor (Neon, running on AWS). Keys are managed by Neon's KMS integration, rotated annually. Encryption is end-to-end: at rest, in transit, and in backup.
Two layers of recoverability. First, every prospect delete and bulk-action delete inside the app goes to a 30-day soft-delete trash with one-click restore. Second, the database itself runs continuous point-in-time recovery: if something corrupts data at a scale the trash can't handle (bad mass-edit, account compromise, corrupted import), we can rewind the entire database to any state within the last 30 days. Two independent layers, neither depends on the other.
Argon2id passwords, TOTP MFA, SSO on Pro+.
Passwords hashed with Argon2id (memory-hard, side-channel resistant). TOTP MFA available to every account, free or paid. SSO via Google Workspace and SAML for any plan at 5+ seats.
Zero-retention. No model training. Ever.
Pro AI calls hit Anthropic with data_retention: 0 and a signed zero-retention agreement. Your prompts and contact data are not used to train any model, ours, theirs, or anyone else's.
One person on the prod box. That's me.
Tyler is the only human with shell access to production. SSH via hardware key. All actions logged to an immutable audit stream. When we hire engineer #2 in 2026, they will get read-only first; write access only after the audit pipeline is dual-control.
Disclosure within 72 hours, every time.
We've had two incidents in 21 months, both rate-limit slowdowns, no data exposure. Both were posted to the public status page within an hour and a full post-mortem within a week. The post-mortems are still online.
The receipts we won't fake to win your procurement review.
Plenty of vendors put compliance badges in the footer the day they sign the engagement letter. We won't do that. We hold no formal compliance certifications today. We do not claim any. Here's the plain version:
Subprocessors. The whole list.
Every third party we send your data through, named, with the region and what they do. If we add or remove one, the change shows up here within 5 business days.
The Dead Man's Switch.
Solo founder. One person. If something happens to me, your data should not be hostage to a probate court. The repo runs a CI job that polls for commit activity. If no commit lands for 90 consecutive days, two things happen automatically:
# SUCCESSION.MD · public · github.com/dealarena/dealarena # cron: every monday 09:00 EST $ last_commit_age > 90d → publish source MIT → mail S3 export to every customer email on file → post final entry on Substack → stripe: refund all unused prepaid balance
Documented in SUCCESSION.md on our public repo. Two trusted contacts hold the secondary keys (one lawyer in Philly, one ops contact in Austin). They get an email if the switch trips.
Found something? Tell me.
No bug bounty bureaucracy. Email Tyler directly with reproduction steps. Acknowledge within 24 hours, fix within 7 days for criticals, public post-mortem after.
security@dealarena.io
PGP:
0xA4F1 2B89 3E7C 0042
avg ack:
3h