Lakera

service

proprietary SaaS API(self-hosted container on an enterprise license)

note

vendor Lakera acquired by Check Point Nov 2025

source

closed

license

proprietary

Lakera Guard, now sold by Check Point as AI Guardrails, is proprietary, with no published source code or model weights. It runs as a hosted API or as a self-hosted container that needs an enterprise license and registry credentials from Check Point, so self-hosting does not open it up.

• https://docs.lakera.ai/docs/api recorded 2026-09-29

  The API reference says "The public Guard SaaS API uses API keys to authenticate requests" and that "Self-hosted deployments of the Guard API do not use authorization so do not need API keys." Searched the body for open source and Apache: none, and its only GitHub links point to GitHub's own documentation on commit hashes.

• https://docs.lakera.ai/docs/selfhosting.md recorded 2026-09-29

  The self-hosting page says "Self-hosting Check Point AI Guardrails allows organizations to keep all data within" their own infrastructure, that the product "is available as a fully managed" SaaS or self-hosted deployment, and that self-hosters need an AI Guardrails Enterprise license and "Container registry credentials (provided by Check Point)". Searched the body for open source, GitHub and Apache: none.

• https://www.lakera.ai/ recorded 2026-09-29

  Commercial site sold through "Book a demo" with "Learning from 1M+ hackers around the world" as its only large figure; searched the body for open source, GitHub repository and license: no source release or open license is offered.

Lakera publishes no customer count, request volume or revenue figure, and the only large figures on its site count players of its Gandalf game and members of its community, not users of the product. Its reach is inferred from its standing as an established prompt-injection-defense vendor rather than from any published count.

• https://www.checkpoint.com/press-releases/check-point-acquires-lakera-to-deliver-end-to-end-ai-security-for-enterprises/ recorded 2026-09-29

  Check Point's acquisition release says "Lakera is trusted by Fortune 500 enterprises worldwide" without a customer count or usage figure.

• https://www.lakera.ai/ recorded 2026-09-29

  The home page claims "trusted by industry leaders, from Fortune 500 companies to startups" and gives no customer count; its only figures are "Learning from 1M+ hackers around the world" (the Gandalf game) and a Slack community "With over 1,000 members".

The Guard API screens prompts, model outputs, tool calls and tool descriptions for prompt attacks, data leakage, content violations, malicious links and off-policy agent behavior, with custom guardrails on top. Check Point now sells it as AI Guardrails, the runtime layer of its AI Agent Security product, which adds agent discovery and risk assessment.

• https://docs.lakera.ai/guard.md recorded 2026-09-29

  The AI Agent Security introduction says AI Guardrails provides "real-time detection and flagging of prompt attacks, data leakage, content violations, and off-policy agent behavior through the Guard API" and screens for malicious links and custom threats, with the runtime layer "available standalone for teams that want to embed Check Point's detection directly into their own AI applications."

• https://www.lakera.ai/ recorded 2026-08-13

  AI Agent Security is described as 'Runtime protection for AI applications' with 'Real-time threat detection', 'Prompt attack prevention' and 'Data leakage protection'; the risks listed are prompt injection attacks, AI data leaks, toxic content generation, indirect prompt injection and multilingual/multimodal attacks, at 'sub-50 ms runtime latency'.