No plaintext access
The engine is architecturally separated from vault contents. There is no API that returns your secrets to it.
ARKCRON AI Security Engine
AI that protects your secrets — without reading them.
A risk engine and security assistant that works on privacy-preserving signals. It never needs unrestricted access to your plaintext vault.
Risk assessment · sign-in
0.87
risk score · block & verify
- New device fingerprint+0.31
- Impossible travel, 2,400 km in 40 min+0.42
- Typing cadence deviates from baseline+0.18
- Known IP reputation−0.04
Every decision lists the signals that produced it.
Privacy model
Analysis where the plaintext is. Intelligence where the context is.
Anything that needs your secrets is computed on your device. The cloud engine only sees what it needs to judge risk.
- On device: strength scoring, reuse detection, credential age, site matching.
- Breach checks: only a short hash prefix leaves the device (k-anonymity); the match is confirmed locally.
- Cloud engine: scores, counts, device and session telemetry — never item contents.
- You decide: recommendations are advisory; enforcement applies only to sign-in policies you or your admin configure.
- 01
Vault
Plaintext exists only on your device.
- 02
Local security analysis
Strength, reuse, age and phishing checks run locally.
- 03
Privacy-preserving signals
Scores, counts, hash prefixes and session telemetry.
- 04
ARKCRON AI Risk Engine
Correlates signals with threat intelligence and behavior baselines.
- 05
Risk score
A calibrated score with the reasons behind it.
- 06
Security recommendation
Rendered on your device. You choose what to do.
Signals
What the engine watches for.
- Credential exposure
- Weak passwords
- Password reuse
- Compromised credentials
- Suspicious login behavior
- Impossible travel
- New devices
- Authentication anomalies
- Phishing indicators
- Identity risk
- Abnormal access behavior
- Credential age
Credential exposure
ARKCRON detected that this credential may have appeared in a recent breach. We recommend replacing it.
Login blocked
Authentication behavior differs significantly from your normal pattern.
Upgrade available
3 accounts support passkeys. Migrating removes their passwords from phishing risk.
Progress
Your overall identity exposure decreased by 18% this month.
Principles
Rules the AI is built to follow.
Explainable decisions
Every score lists its contributing signals, visible to you and, for business plans, to your admins.
Human in control
Recommendations never change your vault on their own. Sign-in enforcement follows your configured policy.
Logged and reviewable
AI decisions about sign-ins are written to the audit log for review and SIEM export.
No training on your vault
Vault contents are never available to train models — they never reach the service unencrypted.
Tunable sensitivity
Enterprises can set risk thresholds per app, group and data sensitivity.
Security intelligence, without surrendering your secrets.
Free for individuals. No card required. Your keys stay on your devices from the first minute.