Developer Tools·Standings

AI Security Testing Tools

AI security testing tools help teams check AI applications for issues such as prompt injection, jailbreaks, data leakage and unsafe output. This board covers 29 tools and shows the test areas, deployment modes and starting prices listed for each one.

Use the signal score out of 10 as a quick comparison. The board also marks whether a tool has a public API, open-source code or a free tier. Twenty-four tools have a free plan; one has a free trial.

  • 29 platforms
  • 2 with a public API on this page
  • 24 with a free tier
  • Checked 23 Sept 2026
  1. 1
    P
    Promptfoopromptfoo.dev

    Tool for comparing prompts across 60+ model providers and simulating AI risks.

    7.7 APIOpenFree Free tier
  2. 2
    AR
    AIRTA Red Teamairtasystems.com

    Free tool that runs authorized attacks against LLMs and applications through API or browser.

    7.7 APIOpenFree Free tier
  3. 3
    OS
    OpenSecureAI Scanneropensecureai.com

    Open-source scanner that detects prompt injection, jailbreaks, leaks, and exfiltration.

    7.7 APIOpenFree Free tier · from $49/mo
  4. 4
    C
    CyberSecEvalmeta-llama.github.io

    Free benchmark suite covering cyber risks, secure code, refusal rates, and prompt injection.

    7.6 APIOpenFree Free tier
  5. 5
    G
    Giskardgiskard.ai

    Platform that finds agent issues such as prompt injection, leaks, hallucinations, and omissions.

    7.4 APIOpenFree Free tier
  6. 6
    P
    PromptGuardpromptguard.co

    Security layer before an LLM that flags threats and masks 43 types of personal data.

    6.7 APIOpenFree Free tier · from $19/mo
  7. 7
    PM
    Project Moonshotaiverifyfoundation.sg

    Open-source toolkit using benchmarks and adversarial prompts to assess LLM safety.

    6.7 APIOpenFree Free tier
  8. 8
    R
    RAMPARTmicrosoft.github.io

    Open-source pytest framework that runs agent attacks or probes and reports outcomes.

    6.3 APIOpenFree Free tier
  9. 9
    A
    Augustusgithub.com

    Go-based scanner with 210+ probes, 47 attack categories, and 28 provider types.

    6.3 APIOpenFree Free tier
  10. 10
    B
    Basiliskbasilisk.rothackers.com

    Open-source red-team framework with 32 modules, prompt evolution, and many provider connections.

    6.3 APIOpenFree Free tier
  11. 11
    P
    PyRITazure.github.io

    Open-source framework for automated or human-led red teaming of generative AI systems.

    6.3 APIOpenFree Free tier
  12. 12
    R
    RedForgegithub.com

    Open-source LLM security platform with 49+ probes and OWASP LLM Top 10 coverage.

    6.2 APIOpenFree Free tier
  13. 13
    P
    PromptRAIDgithub.com

    Open-source framework generating prompt-injection variants and scoring agent responses.

    6.1 APIOpenFree Free tier
  14. 14
    G
    garakgarak.ai

    LLM scanner probing hallucinations, leaks, injection, toxicity, misinformation, and jailbreaks.

    6.1 APIOpenFree Free tier
  15. 15
    S
    SuperRedrdi.berkeley.edu

    Open-source framework coordinating attackers, targets, and benchmarks for chatbot red teaming.

    6.0 APIOpenFree Free tier
  16. 16
    A
    AgentDojoagentdojo.spylab.ai

    Open-source environment for evaluating prompt-injection attacks and defenses in agents.

    6.0 APIOpenFree Free tier
  17. 17
    RT
    Red Team LLM Labgithub.com

    Open-source framework for injection, jailbreak, backdoor, RAG poisoning, and defense checks.

    5.9 APIOpenFree Free tier
  18. 18
    A
    AgentScanagentscan.sh

    API that scans LLMs, RAG pipelines, and agents across 412 attack vectors.

    5.9 APIOpenFree Free tier · from $29/mo
  19. 19
    L
    LLMMapgithub.com

    Linux tool that finds HTTP prompt-injection points and checks 227 techniques with LLM judges.

    5.9 APIOpenFree Free tier
  20. 20
    I
    InjectLabgithub.com

    Open-source YAML test suite covering LLM ATT&CK tactics with scoring and mitigation guidance.

    5.8 APIOpenFree Free tier
  21. 21
    H
    HarmBenchgithub.com

    Open-source tool for comparing automated attack methods and model defenses.

    5.8 APIOpenFree Free tier
  22. 22
    A
    AIRTaaSairtaas.ai

    Web platform for continuous security checks and policy validation of agentic applications.

    5.8 APIOpenFree Free tier
  23. 23
    C
    Cruciblecrucible-security.github.io

    Agent security tool with 153 attack vectors and reports on tool calls and side effects.

    5.8 APIOpenFree Free tier
  24. 24
    D
    DeepTeamtrydeepteam.com

    Open-source framework with 50+ vulnerabilities and simulated attacks for LLM systems.

    5.7 APIOpenFree Free tier
  25. 25
    SR
    SENTINEL REDsentinelred.dev

    Open-source suite covering injection, hallucinations, leaks, poisoning, and compliance.

    5.3 APIOpenFree —

Places #1, #2, #3, #4, #5 on this board are open to sponsors. Take #1 · Other ways to advertise

How to choose

Match the checks to your needs

The board compares prompt injection, jailbreak, data leakage and unsafe output tests, along with custom test cases. Check which of these areas matter for your application before comparing prices.

Custom test cases can help when your risks do not fit the standard checks shown on the board.

Start with the available plan

Twenty-four of the 29 tools have a free plan. One tool lists a free trial, so check the plan details before committing.

Published paid starting points vary. PromptGuard lists paid plans from $19 per user each month, AgentScan from $29 per month and OpenSecureAI Scanner from $49 per month.

Check where it runs

The listed platforms include Linux, self-hosted, API, web, Windows and macOS. Counts are 21 for Linux, 19 self-hosted, 18 API, 14 web, 13 Windows and 11 macOS.

Choose a deployment mode that fits your existing setup. An API option means the software can be connected through a public programming interface; self-hosted means you run it yourself.

Review the code and access model

The board marks open-source projects separately from free plans and free trials. Project Moonshot, CyberSecEval, PyRIT and AgentDojo list a free plan, while RedForge and DeepTeam list open-source pricing notes.

A free plan does not necessarily mean the project is open source, so compare both marks when control over the code matters.

Questions

How many tools are on this board?

The list contains 29 AI security testing tools.

How many offer a free plan?

Twenty-four of the 29 tools have a free plan.

Which checks are compared?

The board compares prompt injection, jailbreak, data leakage and unsafe output tests, plus custom test cases and deployment mode.

Which platforms are listed most often?

Linux appears on 21 items, self-hosted on 19, API on 18, web on 14, Windows on 13 and macOS on 11.

Are open-source tools identified?

Yes. Open-source status is marked separately from free plans and free trials.

Some top positions on this board can be paid placements, standing where they were sold. The facts on each platform are not for sale. How we rank · Disclosure