Every request gets a verdict.
Every verdict gets a signature.
Specter classifies traffic at the edge — human, automated, adversarial — and serves graduated access to automated agents. Every decision ships with a cryptographically signed audit trail proving no real human was ever affected.
Trust earned in three phases.
Nothing turns on until it's proven.
Specter deploys in a staged lifecycle. You see the classification and the false-positive evidence before any graduated serving goes live.
Observe
Classify every request in read-only mode. Measure the false-positive rate against ground truth. Nothing changes for visitors.
✓ zero behavior changeShadow
Review what would happen in a review queue. Approve or override each decision before it affects any traffic.
✓ human in the loopActive
Serve graduated access to automated traffic, with a signed evidence package for every single decision.
✓ earned, not claimedDetection is table stakes.
The proof is the product.
Detect-and-block
- Blocks bots, but attackers rotate IPs and nothing is proven
- Static obfuscation — LLMs read around it
- No attribution, no audit trail
- No way to tell Sales "we never touched a real prospect"
Specter
- Graduated access for automated traffic — never a hard block
- Behavioral classification at the edge, before origin
- Signed EATF evidence package for every decision
- One audit trail satisfies the VP of Sales and the EU auditor
Measured in observe mode with design partners, at active thresholds. Zero documented cases of a legitimate human served graduated content.
"You can't hire the insight of someone who built both the attack and the defense." — the founder's edge, stated plainly
- EU-native. Built for the European surface Estonia has defended since 2007.
- EATF contributor. The evidence backbone is open tooling we already shipped.
- No shared ruleset. Per-customer baselines — deploying Specter leaks nothing about how to defeat it.
- Tell-first. We disclose before anyone discovers. No surprises, no spin.