The AI Security Tool Landscape: 237 Mapped

Real-time AI security. Detects prompt injection, jailbreaks, PII leaks, toxic content. Model-agnostic.

Programmable guardrails. Colang configuration language for custom safety logic.

Toolkit for securing LLM interactions. Modular input/output scanners.

AI agent supervision platform. Runtime monitoring and control of agent actions. Note: different from MITRE Caldera (Apache 2.0, adversary emulation).

Autonomous offensive security -- expert-level pentesting at machine speed without human intervention.

AI-powered API and application security testing discovering business logic flaws.

First comprehensive agentic AI pentesting platform for web apps with human supervision.

AI-powered pentesting toolkit using LLMs as interactive reasoning copilot.

Automated security validation -- full pentesting lifecycle from discovery to exploitation.

AI offensive platform continuously mapping attack surfaces and autonomously exploiting vulns.

AI pentesting MCP framework with 150+ security tools and 12+ autonomous agents.

AI pentesting platform specifically testing AI/ML models against adversarial attacks.

Industry-standard web app scanner now with AI-enhanced vulnerability detection. Community free; Pro/Enterprise paid.

Scanner with 10K+ YAML templates and AI-assisted template generation. Open-source core; cloud paid.

Enterprise vuln management covering 65K+ CVEs with AI module for discovering shadow AI.

Developer-first security covering code/containers/IaC + AI models and agents.

All-in-one DevSecOps combining SAST, DAST, SCA, container scanning with AI prioritisation.

Cloud vulnerability scanner with AI-assisted prioritisation for teams without dedicated security.

BAS platform continuously validating security controls against real-world attack scenarios.

Automated adversary emulation with AI-driven security control validation.

CTEM platform combining BAS, ASM, and automated red teaming with AI optimisation.

BAS simulating attacks across full kill chain with 30K+ scenario playbook.

Adversary emulation platform with AI-assisted campaign construction.

Visual link analysis and OSINT with AI-powered relationship mapping across 100+ sources. Community Edition free.

Automated OSINT from 200+ data sources. Open-source core; SpiderFoot HX cloud paid.

Internet-wide scanning with AI for asset discovery and exposure monitoring.

Search engine for internet-connected devices with ML device fingerprinting.

AI-powered offensive security research accelerating vulnerability research and exploit dev.

Google Project Zero AI that autonomously found real-world zero-day vulnerabilities.

World's first bug bounty platform exclusively for AI/ML. 17K+ researchers, 2,520+ CVEs disclosed.

Leading bug bounty with AI-accelerated triage and researcher matching.

Crowdsourced security with AI matching vulnerability types to specialised researchers.

Industry-leading SIEM with agentic AI Triage Agent and AI Playbook Authoring.

Cloud-native SIEM/SOAR with Copilot for Security for natural-language hunting.

Unified SIEM/SOAR/TI platform with Gemini AI and Mandiant intelligence.

AI-native SIEM with Purple AI for natural-language security analysis.

AI-driven SIEM with best-in-class UEBA for insider threat detection.

AI-driven next-gen SIEM investigating multiple alerts simultaneously.

AI-native SecOps combining SIEM, NDR, Open XDR for MSSPs.

AI assistant across Microsoft security portfolio. NL investigation, autonomous agents for phishing/DLP/identity.

AI-first hyperautomation with Socrates "omni-agent" as autonomous SOC analyst.

Autonomous AI SOC -- 100% alert coverage, 95% triaged in under 2 minutes.

Forensic AI SOC using code-level DNA analysis, endpoint forensics, and reverse engineering.

First AI SOC analyst replicating elite analyst techniques. Deploys in 30 minutes.

Agentic AI SOC Analyst with explainable reasoning. Strong MSSP focus.

Agentic SOC covering full lifecycle from L1 through L3.

AI SOC scaling existing teams with agentic AI.

Industry-leading XDR with Charlotte AI for autonomous cross-platform investigation.

Autonomous EDR/XDR using behavioural AI -- no signatures needed.

AI-driven SecOps unifying SIEM + SOAR + XDR + ASM in single platform.

Self-learning AI detecting novel threats via unsupervised ML -- no rules or signatures.

AI-driven detection with patented Attack Signal Intelligence. 80%+ alert noise reduction.

Unified XDR across endpoints, identity, email, cloud with AI attack disruption.

Behavioural AI detecting sophisticated social engineering and BEC that bypasses traditional filters.

Enterprise email security with Nexus AI for multi-layered threat detection. 99.99% stop rate.

AI email security with transparent, customisable, open detection engine. Core free/self-host; Enterprise paid.

Self-learning NDR creating bespoke "pattern of life" for every device/user.

NDR with Attack Signal Intelligence across network, cloud, and identity.

Cloud-native NDR with real-time SSL/TLS decryption and AI asset discovery.

Open NDR on Zeek with AI-enhanced detection and deep forensic evidence. Zeek OSS free; platform paid.

Agentless CNAPP with security graph + AI CNAPP securing AI workloads in cloud.

Comprehensive CNAPP with AI-driven risk prioritisation and auto-remediation.

Agentless cloud security with GenAI-powered analyst for investigations.

AI cloud security with Offensive Security Engine simulating real attack paths.

AI-powered threat intelligence across open/deep/dark web with predictive capabilities.

Mandiant expertise + Google visibility + Gemini AI analysis.

AI threat intelligence integrated into Falcon with automated attribution.

300+ threat feeds with AI correlation and enrichment.

AI identity security unifying ITDR, PAM, and identity protection.

Agentless identity security extending MFA to any resource including legacy.

AI red teaming toolkit supporting multi-modal adversarial testing.

40+ attack simulations covering OWASP LLM Top 10 + built-in guardrails.

AI security CLI for red-teaming and evaluating LLMs. 127 Fortune 500 customers.

LLM safety tools including CyberSecEval for insecure code detection.

Comprehensive adversarial ML library with 15+ attack categories.

"Nmap for LLMs" -- probes for hallucination, data leakage, prompt injection, jailbreaks.

Fuzzing framework for LLM applications with CI/CD integration.

AI model testing for vulnerabilities, biases, and regressions. OSS library free; Hub enterprise paid.

Scans 35+ ML model formats for malicious code and backdoors. Acquired by Palo Alto Networks.

End-to-end AI security -- supply chain, model scanning, runtime protection. Gartner Cool Vendor.

AI firewall + algorithmic red teaming + model assessment. Formerly Robust Intelligence.

Inference-time protection with AI Model Security Leaderboard. Acquired by F5.

AI governance mapping to EU AI Act, NIST AI RMF, ISO 42001.

Open-source LLM firewall — a self-hosted proxy between your app and OpenAI/Gemini/Claude that blocks prompt injection, redacts PII, and enforces custom policies in real time.

Python library and REST API scanning LLM prompts and responses for injections and jailbreaks — layered YARA signatures, transformer classifiers, and vector-similarity detection. Alpha / research status.

Enterprise GenAI security platform — prompt-injection defence, data-leak prevention, and shadow-AI visibility across employees, applications, and agents.

Edge-layer firewall for LLM applications — discovers AI endpoints and detects prompt injection and PII in requests before they reach your model.

GenAI security platform covering employees, applications, and agents — shadow-AI discovery, DLP for LLM traffic, and jailbreak protection.

Zero-touch GenAI data protection — discovers unapproved AI tools and masks or redacts sensitive data in prompts in real time before it reaches the model.

Agentless, network-level observability of employee AI activity — captures prompts and responses, applies policy, and governs agentic AI at runtime.

Data-lineage platform tracing how sensitive data flows into AI tools — unifies DSPM, DLP, and insider risk into one AI and data security platform.

Cloud-native GenAI DLP stopping sensitive data being pasted or uploaded into AI chatbots across browser, SaaS, and endpoints, plus shadow-AI discovery.

Data-security-first platform (DSPM, classification, blast radius) extended with AI inventory, shadow-AI discovery, posture management, and runtime guardrails.

DSPM for AI — discovers where sensitive data lives across cloud and SaaS and assesses whether and how it can flow into AI tools and workflows.

Data security and governance for Microsoft estates — sensitivity labels, DLP, and Copilot AI usage visibility and controls across Microsoft 365.

SSE platform with GenAI app security — discovers sanctioned and shadow AI apps and enforces real-time, granular usage and data policies inline.

Zero-trust platform with AI asset management — inventories the organisation's AI footprint including AI embedded in SaaS, with inline prompt inspection and DLP.

Agentless browser-extension security turning any browser into a governed workspace — shadow-AI visibility and real-time prompt-level DLP.

Identity-centric SaaS security discovering shadow AI by mapping AI tools and integrations to human and non-human identities across the SaaS estate.

Identity-driven SaaS and AI discovery and governance — finds AI apps via identity and access data, then enforces SSO, offboarding, and access revocation.

AI security posture management in a CNAPP — discovers AI models, pipelines, and shadow AI across clouds, then prioritises misconfigurations via attack-path analysis.

Palo Alto's end-to-end AI security platform — model scanning and posture management plus runtime defence, automated AI red-teaming, and AI agent security.

Dedicated AI governance platform — risk assessments, policy packs, auto-generated model cards, and third-party AI vendor risk management.

Privacy and GRC platform with AI governance — unified AI asset inventory, automated intake, risk assessment and approval workflows, continuous oversight.

Enterprise AI lifecycle governance — bias and drift detection, explainability, and regulatory applicability assessment including the EU AI Act.

Compliance automation with continuous control monitoring and framework support including the EU AI Act and ISO 42001.

Compliance automation for the EU AI Act — automated evidence collection, real-time control monitoring, and cross-framework mapping.

Open-source toolkit of metrics and algorithms to detect and mitigate bias in ML models and datasets across the AI lifecycle.

Open-source Python toolkit to assess and improve fairness of ML models, with mitigation algorithms and disparity dashboards.

Multimodal deepfake detection API and platform scoring video, audio, and images in real time across contact-centre, KYC, hiring, and exec-comms workflows.

Multilayer forensic analysis of video, image, and audio deepfakes via web app, API, and SDK, with sub-second verdicts and KYC liveness integration.

Voice and video fraud detection flagging cloned or synthetic audio in live calls and meetings, built for contact centres and enterprise comms.

Developer-friendly APIs detecting AI-generated or manipulated images, video, and audio at scale, with confidence scores for moderation and fraud pipelines.

Real-time video deepfake detector reading photoplethysmography blood-flow signals in facial pixels to verify a live human on camera in milliseconds.

Content authentication and deepfake forensics protecting enterprises from social engineering, voice scams, and impersonation in high-stakes workflows.

Identity verification platform whose liveness engine flags deepfaked selfies and manipulated media during KYC onboarding worldwide.

Biometric liveness verification providing genuine-presence assurance, blocking presentation attacks and deepfake injection in face-based identity checks.

Automated ID document and face verification with deepfake and synthetic-identity detection built into high-volume onboarding flows.

Synthetic-voice detection for live calls and recorded audio — public basic version plus a higher-accuracy commercial API tier.

Cross-modal deepfake detection API scanning selfies, video, and audio in real time for identity fraud and spoofing.

Digital risk platform combining deepfake and impersonation detection with threat intel, brand monitoring, and dark-web visibility for early warning.

Free deepfake video scanner (web, API, and open-source SDK) analysing uploaded files or URLs for facial manipulation and lip-sync anomalies.

Academic open platform bundling many state-of-the-art detection models for images, video, and audio behind one upload-and-analyse interface.

Two-stage LLM triage (falsifier + red-team pass) of Nuclei JSONL findings via OpenAI-compatible endpoints (vLLM/Ollama).

YAML-driven taskflow agent framework for triaging CodeQL/SAST alerts and filtering false positives.

Code-canary decoys to triage AI-hallucinated ("slop") vulnerability reports flooding bug-bounty programs.

Minimal three-stage LLM pipeline (context → scan → skeptical triage) for zero-day discovery in C/C++.

Generates, tests, and refines Sigma rules from real logs with false-positive checking.

Full-stack AI red-teaming platform covering OpenClaw security scan, agent scan, skills scan, MCP scan, AI-infra vulnerability scan, and LLM jailbreak evaluation.

Security scanner for AI-agent skills used by Claude Code, Codex CLI, Gemini CLI, and similar ecosystems; combines static analysis, AST/YARA/taint checks, optional LLM semantic review, MCP least-privilege/tool-poisoning checks, risk scoring, and SARIF/JSON/Markdown output.

Security scanner for AI agents, MCP servers, and agent skills; the successor path for the original Invariant Labs mcp-scan work.

Security scanner for AI-agent configurations, MCP servers, hooks, and tool permissions with CLI, GitHub Action, and app workflows.

Scanner for agent skills combining YAML + YARA patterns, LLM-as-a-judge, and behavioral dataflow analysis (Codex / Cursor skill formats).

Scanner for MCP servers and agentic tool surfaces, covering tools, prompts, resources, package risk, malware indicators, and deployment readiness.

CLI security scanner for agentic workflows (LangGraph, CrewAI, n8n, etc.) — maps tools/data flows and flags risks.

Security checklist for MCP clients, servers, multi-MCP deployments, lifecycle controls, authz/authn, isolation, and crypto-specific MCP integrations.

Large community cybersecurity skill library for AI agents, mapped to MITRE ATT&CK, NIST CSF, MITRE ATLAS, D3FEND, and NIST AI RMF.

Claude Code / agent-skill bundle for authorized bug hunting and external red-team workflows across web, API, identity, cloud, recon, reporting, Burp MCP, slash commands, and the cbh CLI.

Benchmark environment for prompt-injection attacks and defenses in tool-using LLM agents.

Platform for running MCP servers in isolated containers with per-request identity/access policy, registry and gateway workflows, audit logs, Kubernetes operator support, and observability hooks.

AI-agent firewall and verifiable egress-control layer mediating HTTP, WebSocket, CONNECT, MCP, and A2A traffic to detect prompt injection, secret exfiltration, SSRF, and suspicious outbound actions.

Credential gateway and encrypted vault for AI agents; injects real API credentials at the gateway so agents only see placeholder keys.

Local-first, microVM-backed programmable sandboxes for AI agents with SDKs, CLI, MCP support, and rootless hardware isolation.

Enforcement and evidence layer for agentic deployments: static CodeGuard checks, sandboxing, registry ingestion with SSRF guards, and audit/observability.

Security skill suite for OpenClaw-family agents; AGPL-3.0 licensed.

Flagship machine-learning security library for evaluating and defending models against evasion, poisoning, extraction, and inference attacks across major ML frameworks.

Classic Python toolbox for generating adversarial examples and benchmarking robustness of PyTorch, TensorFlow, and JAX models.

Scans ML model files for unsafe serialization patterns and embedded code, with a focus on model serialization attacks.

Python pickle decompiler, rewriter, and static analyzer for inspecting and detecting malicious pickle/PyTorch payloads.

AI-first security scanner for AI/ML repos, agents, and MCP surfaces; AGPL-3.0 licensed.

OpenAI Gym environment for reinforcement-learning agents that mutate PE malware to evade static ML malware detectors.

Historical "Metasploit for machine learning" framework for experimenting with adversarial robustness of ML models.

CLI for detecting malicious PyPI, npm, Go, RubyGems, GitHub Actions, and VSCode extension packages using Semgrep rules and package-metadata heuristics.

Sandboxed static/dynamic analysis pipeline for open-source packages, capturing filesystem, process, and network behavior and publishing data for malicious-package research.

Fully autonomous multi-agent pentest framework with Docker sandboxing.

Modular, bug-bounty-ready agent framework supporting 300+ LLM models. MIT for research; separate commercial license for production/on-prem.

Autonomous "AI hackers" that dynamically run code and validate vulnerabilities with PoCs (Apache-2.0).

Minimal (~50 LOC) research framework for LLM-driven Linux priv-esc and web pentesting (FSE'23).

AI pentesting CLI assistant with local-LLM support (Llama-3.1, Mistral, DeepSeek).

Official Burp Suite extension exposing Burp to AI clients through MCP.

Offensive-security MCP server with 200+ wrapped tools, specialist agents, and OWASP-oriented probes for authorized testing.

Collection of Claude Code offensive-security subagents for authorized penetration-testing research.

Autonomous AI penetration-testing platform that orchestrates specialized web, AD, Kubernetes, CMS, and framework agents through an MCP-controlled Docker toolbox with local privacy-tokenization for sensitive target data.

Autonomous offensive-security meta-harness that wraps local or API-backed coding agents into a multi-agent recon-to-exploit workflow with MCP/API, War Room UI, tool arsenal, and committed benchmark artifacts.

White-box autonomous AI pentester with strong XBOW-benchmark results. *Shannon Lite is AGPL-3.0; Shannon Pro is commercial.*

Black-box AI pentest framework with MCP, multi-agent spawning, and persistent sessions.

Swarm-intelligence multi-agent pentest with stigmergic blackboard coordination (Go).

LLM offensive-security toolkit.

Learns and ranks regex-like naming patterns from known subdomains to generate likely new candidates.

Convolutional neural network that classifies pentest/recon screenshots (login pages, webapps, old-looking sites, parked domains, and custom 404s) for attack-surface triage.

Machine-learning-assisted web intelligence tool that fingerprints products, versions, CVEs, login pages, debug messages, and related web-server signals from HTTP responses.

AI wrapper around the ffuf web fuzzer that suggests file extensions and paths from the target URL and headers using OpenAI or Anthropic models.

WGAN that learns password distributions from leaks to generate guesses; historical reference implementation of the PassGAN paper (MIT).

DistilBERT text-classification model for email and URL phishing detection, trained on a public Hugging Face phishing-email dataset.

Zero-shot vulnerability discovery in Python repos via LLM call-chain analysis; credited with a 0-day RCE in Ragflow.

Agent-powered security harness for scanning large codebases with coding agents, resumable parallel runs, custom matchers, and optional revalidation.

Official Claude-based semantic SAST GitHub Action that reviews PR diffs.

Neurosymbolic SAST combining LLMs with CodeQL for Java vulnerability detection (MIT).

Agent skills that turn AI coding assistants into a multi-agent SAST scanner.

LLM-driven fuzz-harness generation for OSS-Fuzz; reported 26 real vulnerabilities (incl. CVE-2024-9143 in OpenSSL).

LLM-guided protocol fuzzing extending AFLNet (NDSS'24).

System-prompt hardening fuzzer; 16 attacks × 16 providers.

Automated LLM fuzzer for jailbreaks/prompt injection.

Prompt-injection scanner for custom LLM applications in white-box and black-box modes; GPL-3.0 licensed.

Streamlit app: AI summaries, mindmaps, IOC/TTP extraction, and ATT&CK Navigator layers.

LLM + OCR IOC extraction (pulls IOCs from images/PDFs).

Extracts/aggregates IOCs from feeds; integrates with MISP/ThreatKB (pairs well with LLM post-processing).

Explains imported Windows APIs in PE files via GPT and maps to MITRE ATT&CK.

MCP server (ORKL) for querying the ORKL threat-intel API.

LangGraph SOC automation agent with MCP integrations for Wazuh, Cortex, TheHive, and MISP plus mock-agent test lab.

Open-source AI SOC with readable Python agents, Markdown playbooks, and MCP integrations for triage, investigation, hunting, response, reporting, and forensics.

Agentic SOC platform (LangGraph/Dify) with local-LLM support.

pySigma wrapper and LangChain toolkit for automatic Sigma rule creation and translation; LGPL-2.1 licensed.

LLM-driven incident-response scenario generator using MITRE ATT&CK + ATLAS.

IDA Pro plugin: GPT adds comments and meaningful variable names.

MCP bridge for IDA Pro exposing decompile, disassemble, xref, rename, and debugging workflows to LLM clients.

MCP server exposing Ghidra reverse-engineering ops to any MCP-capable LLM.

Ghidra-focused reverse-engineering assistant with MCP support, Claude Skills integration, and long-form analysis workflows.

Native Ghidra MCP extension with broad tool coverage, headless support, and security-sensitive tool gating.

Ghidra MCP server with large tool coverage, GUI plugin, headless server, and lazy tool loading.

Research project for binary-to-C decompilation with LLMs; code is MIT, but model weights use a more restrictive license.

Burp Suite extension integrating GPT for passive scanning.

Agentic LLM vulnerability scanner and AI red-team kit for jailbreaks, prompt injection, fuzzing, and API stress testing.

Python framework for adding input/output guards, validators, structured-output controls, and Guardrails Hub checks to LLM applications.

LLM monitoring toolkit extracting safety/security signals such as jailbreak similarity, prompt-injection similarity, hallucination checks, PII patterns, toxicity, and refusal metrics.

Archived prompt-injection detector (heuristics + LLM + vector DB + canary tokens).

ML/AI penetration-testing automation tool.

CTFd-based AI red-team training challenges.

Framework for building and testing adversarial jailbreak prompts.

Python framework for adversarial attacks, data augmentation, and training for NLP models; useful for robustness testing beyond chat-only LLM scanners.

Research framework for red-teaming LLMs with auto-generated jailbreak prompts.

ICML 2024 standardized evaluation framework for automated red-teaming and robust-refusal benchmarking.

Canonical Greedy Coordinate Gradient adversarial-suffix attack implementation for transferable attacks on aligned language models.

NeurIPS 2024 open robustness benchmark and leaderboard for generating and defending against LLM jailbreaks.

Original PoC for indirect prompt-injection attacks.

Curated catalog of practical defenses against prompt injection.

Apache-licensed prompt-injection classifier usable via Transformers pipelines and ONNX.

Meta prompt-injection and jailbreak classifier from the Llama Guard family.

Open cybersecurity-tuned LLM family for vulnerability analysis, log/traffic investigation, anomaly detection, attack/defense reasoning, command analysis, and security Q&A.

Defense-focused cybersecurity LLM based on Llama-3.3-70B, trained on an alignment-safe security instruction dataset for SOC, cloud, AppSec, detection, and vulnerability-management workflows.

Cybersecurity-oriented Qwen2.5-Coder fine-tune positioned for offensive and defensive security assistance.

Mistral-7B-Instruct fine-tune for cybersecurity assistance, trained on hand-crafted security and hacking-related instruction pairs.

Low-code honeypot using LLMs to simulate SSH/HTTP/MCP services (Go).

Proof-of-concept LLM-powered SSH honeypot that evaluates sessions as benign, suspicious, or malicious.

LLM-powered SSH honeypot (paper *"LLM in the Shell"*).

Multi-protocol LLM honeypot framework (successor to shelLM).

Cowrie SSH honeypot extended with prompt-injection traps to detect LLM hacker agents.

EnIGMA offensive-CTF mode; SOTA on NYU CTF, InterCode-CTF, and Cybench (v0.7 branch).

Maintained Inspect AI evaluation suite containing multiple cyber benchmarks and tasks.

Deliberately vulnerable LangChain ReAct agent for practicing prompt-injection and Thought/Action/Observation injection attacks.

Claude Code plugin orchestrating recon → vuln classes → reporting.

Real-world AI/ML exploits (Metasploit modules + Nuclei templates) for MLflow, Ray, H2O.

GPT-based discovery of privilege-escalation paths in AWS IAM policies.

Proof-of-concept LLM OSINT framework using knowledge and web agents for internet research workflows.

Curated AI-OSINT dorks, queries, and techniques for discovering exposed LLM and AI infrastructure.

MCP server for dnstwist DNS fuzzing to support typosquatting, phishing, and lookalike-domain analysis.

OpenAI embeddings + Qdrant over OSINT corpora.