Skip to main content

Find & fix LLM vulnerabilities

Open-source vulnerability testing for AI

LLM Vulnerability Scanner
Evaluations used by builders at...
ShopifyDiscordAnthropicMicrosoftDoordashCarvana

Uniquely adapted to your LLM product

No generic scans. Promptfoo customizes vulnerability probes specifically for your application.

Comprehensive Security & Policy Coverage

PII Leaks

Prevent the exposure of personally identifiable information.

Competitor Endorsements

Avoid suggesting competitor products or services.

Unintended Contracts

Don't make promises your business can't keep.

Prompt Injection

Protect against unauthorized prompt manipulations.

Jailbreaking

Prevent users from bypassing restrictions.

Excessive Agency

Limit the model's desire & ability to take unwanted actions.

Overreliance

Prevent reliance on incorrect input assumptions.

Hijacking

Safeguard against unauthorized or off-topic use.

Hallucination

Minimize the generation of false or misleading information.

Detect & fix critical failures

Discover hidden risks and take action to protect your brand and users.

Sample vulnerability report

Continuous monitoring

Keep tabs on your LLM application's security posture across different models, prompts, and applications.

Continuous monitoring dashboard

Why choose our scanner?

Comprehensive, data-driven pentests

Verify security, privacy, and compliance performance for models, fine-tunes, RAGs, and agents.

Custom probes targeting your application

Automatically generate adversarial tests targeted specifically at your application and use case.

Detailed vulnerability reports

In-depth analysis of vulnerabilities and practical remedies.

Open Source

Completely open source with a community of seasoned contributors from major tech companies.

Uncover hidden brand and legal risks