PG PromptGuard Scan

LLM penetration testing

LLM Penetration Testing Workflow

LLM penetration testing looks for practical ways an attacker can manipulate an AI application, not just whether a model can answer a prohibited prompt.

Run demo scan View pricing plans

Where it fits

  • A buyer requests evidence that the AI feature has been tested against prompt injection.
  • A product includes agent tools that can change data or trigger business workflows.
  • A regulated team needs a retestable process before approving AI features.

Operational steps

  • Define scope: models, prompts, tools, retrieval sources, memory, logging, and user roles.
  • Run adversarial tests across direct prompts, indirect content, multi-turn conversations, and tool outputs.
  • Document exploitability, impact, reproduction steps, and server-side control gaps.
  • Retest fixes and keep the test pack in CI so the issue does not return.

Common risks

  • The test stops at model behavior and misses app-level authorization failures.
  • Findings are not mapped to owners or release gates.
  • A fix is made in the prompt but not enforced in backend policy checks.

How PromptGuard Scan fits the workflow

PromptGuard Scan gives penetration testers and product security teams a repeatable harness, useful reports, and checkout-ready plans for ongoing LLM release testing.

Ready to test a real AI surface?

Pricing

Team annual is selected by default.

Annual billing is 50% off. All plans use Polar checkout and keep the product page open.

Dev

For solo builders validating one product before launch.

$25/mo
$294 billed yearly. Save 50%.
5 apps500 scans
  • Prompt injection scans
  • Jailbreak template checks
  • PII and key leak detection
  • HTML risk report
  • Email support
Recommended

Team

For engineering teams shipping AI apps through pull requests.

$75/mo
$894 billed yearly. Save 50%.
20 appsUnlimited scans
  • Everything in Dev
  • GitHub Actions and GitLab CI gates
  • Multi-turn bypass testing
  • PDF export and CVSS scoring
  • Shared workspaces and API access

Enterprise

For platform teams, private deployments, and audit-heavy AI systems.

$250/mo
$2,994 billed yearly. Save 50%.
Unlimited appsUnlimited scans
  • Everything in Team
  • Private deployment path
  • Custom test packs
  • Compliance evidence exports
  • Priority security review support

Security playbooks

Practical guides for LLM app security decisions.

prompt injection scanner Prompt Injection Scanner for LLM Apps Read workflow LLM security testing tool LLM Security Testing Tool for Release Gates Read workflow AI app security audit AI App Security Audit Checklist Read workflow jailbreak detection API Jailbreak Detection API for AI Products Read workflow LLM vulnerability scanner LLM Vulnerability Scanner for Agents and RAG Read workflow AI red team SaaS AI Red Team SaaS for Continuous Testing Read workflow ChatGPT security testing ChatGPT Security Testing for Apps and Custom Workflows Read workflow LLM penetration testing LLM Penetration Testing Workflow Read workflow

Related AI workflow reference

Teams comparing workflow plans with launch and market assumptions can also review MiroFish AI Simulator, a companion reference for simulation-style product reasoning.