Lakera (Check Point)
Guardrail and red teaming product, now sold as part of Check Point
Quick Overview
| Company | Lakera (Check Point Software Technologies Ltd) |
|---|---|
| Category | AI Security Platforms |
| Headquarters | Tel Aviv (Check Point), Israel |
| Founded | Not stated |
| EU Presence | Yes - Israel (counts as European for EP) |
| Data Location | Processing endpoints in the EU, USA and Singapore; storage region configurable |
| Open Source | No |
| GDPR Compliant | Yes |
| Self-Hosting | Self-hosted container available |
| Main Features | Prompt defense, content moderation, data leakage prevention, malicious link detection, agent behavior defense, custom guardrails, allow and deny lists, AI red teaming, agent discovery |
| Pricing | See website (pricing page not readable without scripts) |
| Best For | Teams that want an API guardrail with a regional processing choice |
| Replaces | HiddenLayer AI guardrails |
Detailed Review
Lakera Guard screens what goes into and comes out of an LLM and returns a verdict. The documentation now brands the product Check Point AI Guardrails, and the lakera.ai site links its terms and privacy pages to Check Point.
The Check Point site terms define the company as Check Point Software Technologies Ltd, registered in Israel at 5 Shlomo Kaplan Street, Tel Aviv. That is the entity behind the Lakera pages reviewed, which is why EP lists it as an Israeli-owned product.
Five defences in one API
The docs list prompt defense for jailbreaks and injection, content moderation, data leakage prevention, malicious link detection and agent behavior defense, which flags dangerous deviation and applies a tool allow or deny list.
Custom guardrails and per-project policies let each application carry its own settings.
Choosing the region
Processing and log storage are separate settings. The EU endpoint processes in eu-south-2, and US and Singapore endpoints exist alongside it. The docs warn that routing is not automatic: requests are processed wherever they are sent.
Red teaming and discovery
Beyond runtime, the same platform covers AI red teaming with a Management API and SDK, plus agent discovery and risk assessment. Check Point is the documented vendor for all of these.
Limitations to Consider
Lakera is no longer an independent start-up. Branding in the docs has moved to Check Point, which can mean roadmap and contract terms follow a large vendor. The pricing page needs scripts to render, so no price is stated here. Latency and language figures on the homepage are vendor claims. For an independent European specialist, see NeuralTrust.
Who Should Use Lakera (Check Point)
Teams already using Check Point, or developers who want an API-first guardrail with EU processing and an option to run it in their own container.
Pros and Cons
Pros
- Five guardrail types behind one API
- EU processing endpoint
- Storage region configurable separately
- Self-hosted container available
- Red teaming and discovery on the same platform
Cons
- Now owned by a large security vendor
- No readable public price
- Region routing is the customer's job
- Docs rebranded away from Lakera
Alternatives to Lakera (Check Point)
Other European ai security platforms worth comparing before you decide:
Mindgard
United Kingdom · London AI red teaming that attacks, fixes and retests
NeuralTrust
Spain · Barcelona agent gateway and runtime guard with free tier
Advai
United Kingdom · UK AI assurance: choose, test and monitor AI systems
Giskard
France · Paris open-core red teaming and testing for AI agents
See all European ai security platforms or compare them directly against HiddenLayer AI guardrails on our alternatives page.
Frequently Asked Questions
No. The lakera.ai pages link to Check Point terms and the documentation calls the product Check Point AI Guardrails.
Yes. The EU endpoint processes requests in eu-south-2, but you must send requests to that endpoint yourself.
The docs say AI Guardrails can run as a self-hosted container or through the SaaS service.