AI Defense Stack

Comparison: Guardrails

Lakera Guard vs Google Cloud Model Armor

On Google Cloud, Model Armor screens prompts and responses inside your project with templates and project-wide floor settings; Lakera Guard is a hosted API, also offered self-hosted, that adds system prompt leakage and agent tool defenses. Both are closed source and screen for attacks, sensitive data and harmful content. Lakera's docs now carry the Check Point AI Guardrails name.

  • Last reviewed

Executive summary

What it is
Lakera Guard and Google Cloud Model Armor are closed-source services that screen prompts and model responses for attacks, sensitive data and harmful content. Lakera Guard fits teams that want a hosted or self-hosted API; Model Armor fits Google Cloud projects.
The risk
Unscreened prompts and retrieved content let injected instructions steer the model, leak the system prompt or misuse agent tools, and phishing links in responses reach users.
What to buy or do
Choose Lakera Guard for system prompt leakage, tool allow and deny lists or self-hosting. Choose Model Armor on Google Cloud for templates, floor settings and Apigee integration; Google documents single-turn inspection. Start either in log-only mode before you block.

Which should you choose?

Our analysis, built from the facts in the table below.

Choose Lakera Guard if

Our analysis
  • You want a hosted API that you call for each interaction or agent step, with a self-hosted option for stricter environments.
  • You want system prompt leakage prevention and a tool allow and deny list among the documented defenses.
  • You want to start in Detect mode, where Lakera shows detections in its dashboard without flagging requests, before you enforce a policy.

See Lakera Guard.

Choose Model Armor if

Our analysis
  • You build on Google Cloud and want screening with documented integrations for Gemini Enterprise Agent Platform, Apigee, LangChain and Google and Google Cloud MCP servers.
  • You want templates with project-wide floor settings, plus an inspect-only mode that logs to Cloud Logging before you switch to inspect-and-block.
  • You want Sensitive Data Protection with de-identification in the screening path, which the advanced configuration offers.

See Model Armor.

Side by side

Every cell links to the page that states it, with the date we checked it. We compile from public pages and repositories and do not run either product. Coverage Score is computed within one category from sourced facts (how it works).

Lakera Guard compared with Model Armor: each row shows the fact for both tools with its source
Attribute Lakera Guard Model Armor
Vendor Check Point (Lakera) Source, checked October 21, 2026: checkpoint.com (opens in a new tab) Google Cloud Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Status Renamed Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Active Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Ownership Check Point completed its acquisition of Lakera in the fourth quarter of 2025. Source, checked October 21, 2026: checkpoint.com (opens in a new tab) No ownership change documented in the sources we checked
Open source and license Closed source (proprietary) Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Closed source (proprietary) Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Latest release Not documented in the sources we checked Not documented in the sources we checked
Deployment SaaS, Self-hosted, API Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) API, Cloud-native, Managed service Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Pricing model Not published in the sources we checked Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Usage-based Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
OWASP LLM risks Our analysis
  • LLM01 Prompt Injection
  • LLM02 Sensitive Information Disclosure
  • LLM06 Excessive Agency
  • LLM07 System Prompt Leakage
  • LLM01 Prompt Injection
  • LLM02 Sensitive Information Disclosure
Documented capabilities
  • Detects direct and indirect prompt injection
  • Detects jailbreak attempts
  • Screens content for harmful or unwanted material under the organization's policies
  • Protects personally identifiable information and other sensitive data from leaking
  • Prevents system prompt leakage
  • Flags unknown links that an attacker could get the model to show users
  • Controls which tools an agent can call at runtime with a tool allow and deny list
  • Lets you describe a custom security or content policy in plain language (beta)
Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab)
  • Detects prompt injection attempts that try to bypass system guardrails
  • Detects jailbreak attempts as part of the same filter
  • Detects and can redact PII and other sensitive data with Sensitive Data Protection
  • Checks for hate speech, harassment, sexually explicit and dangerous content at configurable confidence levels
  • Scans prompts and responses for malicious URLs
  • Applies templates and project-wide floor settings that set filters and minimum requirements
  • Logs detections to Cloud Logging in inspect-only or inspect-and-block mode
Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Integrations 1 listed
  • Agent pattern that screens tool definitions, such as at the MCP handshake, and tool inputs and outputs
Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) 		4 listed
  • Gemini Enterprise Agent Platform
  • LangChain
  • Google and Google Cloud MCP servers
  • Apigee API gateway
Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab) Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab) Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Documentation
  • Documentation site
  • API reference
  • Quickstart
Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab)
  • Documentation site
Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
How you call it An API request to the Guard API for each interaction or agent step; the project mode (Detect or Enforce) decides whether a threat is flagged, and the docs suggest blocking or logging Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) REST methods SanitizeUserPrompt and SanitizeModelResponse, with inspect-only or inspect-and-block enforcement Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Where it runs Cloud-hosted SaaS, or self-hosted with configuration through JSON files on S3-compatible storage Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) A managed Google Cloud service, standalone or part of Security Command Center, that the overview says can help protect AI deployed in Google Cloud or other cloud providers Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Defenses listed Prompt Defense, Content Moderation, Data Leakage Prevention, Malicious Links Detection, Agent Behavior Defense and Custom Guardrails (beta) Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Responsible AI safety, prompt injection and jailbreak detection, Sensitive Data Protection and malicious URL detection Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Notes from the docs Self-hosted features are released after an additional stability verification period, while SaaS gets early-access features Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Each prompt and response is inspected as a single turn, encoded content such as Base64 is not decoded, and audio and video are not supported Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Log-only rollout Detect mode is observation only: detections show in the response breakdown and dashboard, and the top-level flagged field stays false Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Inspect-only enforcement logs violations to Cloud Logging for monitoring and analysis but does not stop the request or response Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Link scanning Malicious links detection flags unknown links that an attacker could get the model to show users; the defenses page does not state a link limit Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Scans only the first 256 URLs found in prompts and responses when malicious URL detection is on Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Integrations documented Integration guide covers an AI gateway pattern, chat, document processing, and agent and workflow patterns Source, checked October 21, 2026: docs.lakera.ai (opens in a new tab) Gemini Enterprise Agent Platform, Apigee, Gemini Enterprise, Agent Gateway, Google Cloud networking services, LangChain and Google and Google Cloud MCP servers Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Ownership Check Point completed its acquisition of Lakera AI in the fourth quarter of 2025 Source, checked October 21, 2026: checkpoint.com (opens in a new tab) A Google Cloud service, described as such in Google Cloud's own documentation Source, checked October 21, 2026: docs.cloud.google.com (opens in a new tab)
Coverage Score 57 out of 100 Computed from sourced facts. What this measures 44 out of 100 Computed from sourced facts. What this measures
Where each is stronger Our analysis
  • Documents system prompt leakage prevention and an agent tool allow and deny list; we did not find either in the Model Armor overview.
  • Documents detection of direct and indirect prompt attacks, and custom guardrails in beta where you describe a policy in plain language.
  • Offers a self-hosted version as well as a cloud-hosted service.
  • Documents templates and project-wide floor settings that set filters and minimum requirements across a project.
  • Documents Sensitive Data Protection with de-identification in its advanced configuration.
  • Documents integrations with Gemini Enterprise Agent Platform, Apigee, LangChain and Google and Google Cloud MCP servers.

Verdict

Our analysis

Our analysis

Where you run decides most of this. Lakera Guard fits teams that want a hosted API with a self-hosted option and defenses for system prompt leakage and agent tool use. Model Armor fits teams on Google Cloud that want templates, floor settings and integrations such as Apigee, though its docs say it can also help protect AI on other clouds. Both are closed source and both can start in a log-only mode. Lakera's docs now use the Check Point AI Guardrails name, so confirm current naming. We could not read a price for either product on the pages we opened, and we did not measure detection quality.

Lakera Guard vs Model Armor: FAQ

Are Lakera Guard and Model Armor open source?

No. Both are closed-source services, and the vendors' documentation links no public source repository for either product.

What is the main difference between Lakera Guard and Model Armor?

Lakera Guard is a hosted API, also offered self-hosted, with defenses that include system prompt leakage and an agent tool allow and deny list. Model Armor is a managed Google Cloud service built around templates, floor settings and four filters. The table above shows the source for each fact.

Has Lakera Guard been renamed?

Lakera's documentation now describes the product as Check Point AI Guardrails. Check Point completed its acquisition of Lakera in the fourth quarter of 2025. Check the vendor for current naming.

Does Model Armor keep conversation context?

No. Google's documentation says each prompt and response is inspected as a single turn, without conversation history, and that it does not decode encoded content such as Base64.

Can I run either product without blocking traffic at first?

Yes. Lakera projects can run in Detect mode, which is observation only and shows detections in the dashboard. Model Armor has an inspect-only mode that logs to Cloud Logging without stopping the request or response.

Key facts you can cite

Keep exploring

Sources & last reviewed

Last reviewed . We compile from the vendors' own public pages and repositories and we do not run the products. Source links open in a new tab. Found a mistake? Report a correction.

Drafted with AI assistance; facts checked against primary sources on October 21, 2026.

Lakera Guard (10 sources)
Google Cloud Model Armor (4 sources)
Comparison rows (16 sources)