Loading…
Security · Lakera (Cisco)
Real-time guardrails against prompt injection and jailbreaks for AI apps.
Lakera Guard sits between users and LLMs as a low-latency security layer, detecting and blocking direct and indirect prompt injection, jailbreaks, and system-prompt extraction across 100+ languages. Its models are trained on adversarial data from Gandalf, Lakera's prompt-injection game. Acquired by Cisco in 2025.
Model support
Guards any LLM-backed app via an API in front of the model.
Where it runs
Tags
Related in Security
HiddenLayer
AI security platform: model scanning, runtime defense, and automated red-teaming.
HiddenLayer's AISec Platform unifies AI supply-chain security, runtime detection and response, posture management, and automated red-teaming for generative, agentic, and predictive AI. Its Model Scanner inspects model files for malware and integrity issues, while AI Detection & Response monitors prompts and responses with deterministic classifiers.
AI insight: Its runtime classifiers sit outside the model's inference path, so detection adds no latency to the model itself — unlike inline guardrails.
Lasso Security
End-to-end GenAI security: shadow-AI discovery, red-teaming, and runtime guardrails.
Lasso Security protects every LLM interaction across cloud and on-prem deployments, combining shadow-AI discovery, real-time threat detection, policy enforcement, and red-teaming with thousands of attack types. It deploys via gateway, API, or SDK to secure both internal LLM apps and employee use of third-party chatbots.
AI insight: Lasso frames LLM security as a five-stage lifecycle (Discover, Assess, Test, Enforce, Protect) and sells it as one product, not add-ons.
Pangea
API-based security guardrails for AI apps: prompt injection, PII redaction, and access control.
Pangea provides composable, API-first security services for AI applications. AI Guard and Prompt Guard defend against prompt injection, sensitive-data disclosure, and malicious content, alongside AI access control and visibility products. Pay-as-you-go with a free monthly balance, integrable via LiteLLM and Portkey.
AI insight: Sells security as composable API services (AI Guard, Prompt Guard, audit log, redaction, authz), so you add guardrails with plain API calls.
Prompt Security (SentinelOne)
Runtime security for enterprise GenAI: shadow-AI visibility, data-leak and prompt-injection defense.
Prompt Security secures both employee use of GenAI tools and homegrown LLM applications, giving organizations visibility into shadow AI, blocking sensitive-data leakage, and preventing prompt injection in real time. Acquired by SentinelOne in 2025 and integrated into its Singularity platform.
AI insight: SentinelOne acquired Prompt Security in 2025 (~$250M) to anchor its GenAI/agent-security line within the Singularity platform.
Protect AI (Palo Alto Networks)
End-to-end security for ML models, AI apps, and the AI supply chain.
Protect AI secures the AI/ML lifecycle from model selection through deployment and runtime: model scanning for serialization attacks, AI red teaming, posture management, and runtime protection. It also stewards open-source tools like ModelScan. Acquired by Palo Alto Networks in 2025 and integrated into the Prisma AIRS platform.
AI insight: Acquired by Palo Alto Networks in 2025 and folded into Prisma AIRS; stewards the open-source ModelScan and NB Defense tools.
Snyk
Developer security platform with DeepCode AI for SAST, SCA, and AI-generated fixes.
Snyk finds and fixes vulnerabilities in code, open-source dependencies, containers, and IaC. Its Snyk Code SAST engine, DeepCode AI, combines symbolic and machine-learning analysis with inter-file data-flow tracing to detect issues and auto-generate fixes. Integrates into IDEs, the CLI, and CI.
AI insight: Snyk Code is built on DeepCode AI, the symbolic+ML engine from its 2020 DeepCode buy, trained on millions of data-flow cases, not an LLM.