Skip to main content

Free AI Security Assessment

Submit your LLM endpoint. We run 72 adversarial attack scenarios. You get a detailed security report in 48 hours. No cost, no strings attached.

72-SCENARIO REPORT

Adversarial Test Results

Full red team assessment covering prompt injection, jailbreaking, data exfiltration, privilege escalation, and more. Difficulty-graded from basic to expert.

SHIELD TUNING

Defense Recommendations

Specific threshold adjustments, pattern additions, and configuration changes based on what bypassed your current defenses.

FRAMEWORK MAPPING

OWASP + MITRE Coverage

Every finding mapped to OWASP LLM Top 10, MITRE ATLAS techniques, and CWE identifiers. Ready for compliance documentation.

How It Works

1

Submit Your Endpoint

Fill out the form below with your LLM API endpoint and authentication details.

2

We Run Oubliette Dungeon

72 attack scenarios across all OWASP LLM Top 10 categories, graded by difficulty.

3

Get Your Report

Detailed PDF report delivered within 48 hours with findings, recommendations, and framework mappings.

Request Your Assessment

Your endpoint credentials are handled securely and never stored beyond the assessment period. We will coordinate authentication details via encrypted email before testing begins.

Frequently Asked Questions

Is it really free?

Yes. The assessment uses our open-source Oubliette Dungeon tool (72 attack scenarios). We run it, format the results, and deliver the report at no cost. We believe every AI system deserves a security baseline.

What endpoints can you test?

Any HTTP-accessible LLM endpoint that accepts text input and returns text output. This includes chat completions, RAG pipelines, AI agents, and custom wrappers. We support OpenAI-compatible, Anthropic, and custom API formats.

Is my data safe?

We only send adversarial test prompts to your endpoint -- we never access your training data, system prompts, or user data. All test results are encrypted and deleted 30 days after delivery. We can sign an NDA before testing if required.

What about on-prem or air-gapped environments?

For environments that cannot expose an endpoint, we provide a self-service option: install oubliette-dungeon from PyPI and run the assessment locally. We will help you interpret the results.

pip install oubliette-dungeon

What happens after the assessment?

You get the report with zero obligation. If you want to improve your defenses, we can discuss deploying Oubliette Shield or running a more comprehensive engagement. But the assessment stands on its own.

Why Trust Us

  • Veteran-owned cybersecurity company
  • SAM.gov registered for federal contracting
  • Open-source tools on PyPI and GitHub
  • Apache 2.0 licensed -- read every line of code