LUNAROUTE
Log in / Sign up

AI for cyberdefense

Understand threats. Build defenses.

Planned model access for verified security teams investigating threats, developing detections, and responding to incidents—including AI-enabled attacks. Support for your analysts, not a replacement for their judgment.

Early-access waitlist. Email only to join. We’ll verify your identity before access.

Human-led defense / 002Illustrative workflow
Threat signals flow through AI-assisted analysis and human review before detection and response.Illustrative analyst workflow—not automatic protection or a live monitoring dashboard.THREAT SIGNALSMessagesCodeActivityAI-assisted analysisConnect evidence. Explore hypotheses.Human reviewYour team chooses the next step.DetectionResponse
Models support the investigation. Your team decides.
01

Verified identity

KYC / KYB verification by LunaRoute.

02

Separate terms

Signature required before access.

03

30-day content retention

Full prompts & responses by default. ZDR requires additional approval.

01 / Defense & response

Cyberdefense

Help security teams turn signals into investigations, detections, and response decisions.

Threat analysis

Investigate suspicious code, activity, and indicators to help analysts form and challenge hypotheses.

Detection engineering

Draft and review detection logic, build representative test cases, and examine gaps before changes reach production.

Incident response

Organize evidence, evaluate response options, and support remediation decisions with human review.

Build defenses against AI-enabled threats.

Attackers can use AI to adapt phishing, scale malicious automation, and generate or modify code. Explore these scenarios with model-assisted analysis, test detection strategies, and validate mitigations in authorized environments.

Model access is not a managed detection or automatic protection service. Keep investigation, response, and deployment decisions under human review.

Investigating vulnerabilities or testing systems? Explore Cybersecurity →

02 / The models

More capability.
A different access standard.

Abliterated variants have reduced refusal behavior. Access still comes with verification, validation, and separate program terms.

Model preview01

GLM 5.3 Cyber

Abliterated variant for research and defense in the planned cybersecurity program.

Verified access / terms required
Model preview02

DeepSeek 4.1 Flash Cyber (1070)

Abliterated variant for research and defense in the planned cybersecurity program.

Verified access / terms required

Model examples for the planned program. Joining the waitlist does not grant model access.

03 / The access path

Trust is the entry requirement.

  1. 01

    Join the waitlist

    Leave your email. No identity documents are collected by this form.

  2. 02

    Complete verification

    We will carry out KYC (Know Your Customer) checks for individuals and KYB (Know Your Business) checks for organizations before model access.

  3. 03

    Sign separate terms

    You will need to sign separate terms for the cybersecurity program before model access. These will cover the program’s retention policy.

  4. 04

    Get approved access

    Access is provided after verification, validation, and signing the program’s separate terms.

30-day retention by default.
ZDR with additional approval.

By default, full prompts and model responses are logged for security purposes. Logs are retained for 30 days, then deleted.

Zero Data Retention (ZDR) is available but requires additional approval.

This cybersecurity program is separate from LunaRoute’s standard Zero Data Retention inference offering.

Get on the list.
Go through the right doors.

For researchers and security teams doing authorized cybersecurity research and defense.

Joining the waitlist does not guarantee approval.