No account required · Quick scan in about 10 minutes

Is your AI system secure
against real attacks?

RedLens generates actual adversarial attacks against your AI, analyzes the responses, and delivers a board-ready security report. Free, in minutes, with no setup required.

1
Configure
2
Test
3
Report

Step 1 of 3 Describe your AI system (about 1 minute)

What you get

  • A risk score and a verdict on every attack, with the reason.
  • Plain-English fix guidance for your top findings.
  • A PDF mapped to HIPAA, the NIST AI RMF and the OWASP GenAI Top 10: evidence to support your compliance work, not a certification.
  • A private report link that works for 30 days.

What happens next

  1. Describe your AI system below (about 1 minute).
  2. RedLens attacks a simulation of it and scores each reply (about 1 to 2 minutes), or you test your live system by hand.
  3. Read your report here, download the PDF and copy your private link.
Required: your name, work email, the authorization checkbox and, to simulate, your system prompt or a short description. Test only AI systems you own or are authorized to test. Nothing connects to your live system.

More detail: what the free assessment tests (1-minute read)

RedLens generates real adversarial prompts against the AI system you describe: prompt injection and instruction override, system-prompt extraction, data exfiltration, out-of-bounds and tool-abuse attempts, and containment escapes, drawn from the six categories of the Master Attack Schema. You choose the categories and the depth. By default RedLens sends each attack to a simulation of your assistant (its own model carrying the instructions you give it) and scores the replies; in manual mode you copy each prompt into your own AI, paste its response back, and RedLens scores it. Neither mode connects to your system and no credentials leave your browser, which is why it needs no account and no setup.

The report you download at the end lists every attack with the verdict and the evidence, maps each confirmed weakness to HIPAA Security Rule sections, the NIST AI RMF and the OWASP GenAI Top 10 as evidence to support your compliance work (it does not certify compliance), and ends with prioritised remediation. If you would rather see the shape of a RedLens report first, click through the interactive sample report.

When you want the same testing run automatically against your own endpoint, with history, scheduling and evidence packages, create a free account (25 hosted scans a month, no card). Teams in regulated sectors can read how the platform fits healthcare, law enforcement and enterprise AI deployments.

How do you want to test?
Both are free. Neither connects to your live system.
Tell us about your AI system
This helps RedLens generate attacks that are realistic and relevant to your environment.
What your team calls this AI. Left blank, the report says "My AI assistant".
In simulate mode this is what the stand-in runs. Nothing you paste here is stored beyond this assessment.
Choose what to test
Select the attack types you want RedLens to run. We recommend selecting all that apply to your use case.
AI Instruction Override
Can an attacker make your AI ignore its rules?
Sensitive Data Extraction
Can an attacker pull confidential info out of your AI?
Safety Filter Bypass
Can an attacker route around your AI guardrails?
Agent & Tool Exploitation
If your AI can take actions, can an attacker hijack them?
Memory & Context Manipulation
Can an attacker corrupt what your AI remembers?
AI Containment Escape
Can your AI agent break out of its sandbox?
Get your free assessment
RedLens's own Claude-powered attack engine generates and scores your assessment. No API key, no account, no cost to you.
Add a phone number or company (optional)
We'll use this to follow up about your assessment. At the end you get a private link to your report and a PDF. No spam.

Free · No account needed · Results download as PDF

Running security assessment
For each attack, copy the prompt → send it to your AI → paste the response back → click Analyze.
Generating attack prompts... 0%
Security Assessment Report
- RISK SCORE
-
Attacks Run
-
Vulnerabilities Found
-
Attacks Blocked
What we found
Each vulnerability below is a confirmed security issue. Your AI failed to defend against these attacks.
ResultAttack TypeWhat happenedHow to fix it
Test your real system, automatically
A free account runs these attacks against your own AI endpoint on a schedule, with history and evidence. No card.
Start free Book a 30-minute review
Full attack log
Every attack attempted in this assessment, with your AI's response and our analysis.
Recommended next steps
What to do with these results.
1
Download this report
Share it with your security team and leadership. The PDF includes all findings, evidence, and remediation guidance.
2
Prioritize critical vulnerabilities first
Any finding marked Vulnerable should be addressed before this AI system handles sensitive data or autonomous actions.
3
Set up continuous monitoring with RedLens
This was a point-in-time assessment. A RedLens account re-runs these tests on a schedule, so regressions and new attack techniques show up in your history.
Where to go from here
Pick whichever fits your team.
Test your real system
A free account runs these attacks against an AI endpoint you own or are authorized to test, with history and evidence. No card.
Start free
Talk it through
Walk through these findings with the RedLens team in a 30-minute call.
Book a call
Design Partner Program
Work directly with the RedLens team on your AI systems and help shape what we build next.
Become a design partner
Guided remediation · preview only