Lakera

Lakera is an AI security product family within Check Point. Its technology protects AI applications and agents through guardrails, discovery, risk assessment, and red teaming. The current offering distinguishes the AI systems an organization builds from the AI tools its employees use.

Key Highlights

• AI Guardrails for prompt attacks, data leakage, and policy violations
• Discovery of agents, their tools, and connected MCP servers
• Agent risk ratings and security-framework mappings
• Automated and expert-led AI red teaming
• SaaS access and optional self-hosted AI Guardrails

What Makes It Different

Lakera connects runtime checks with an inventory of agent capabilities. AI Agent Security combines discovery and risk assessment with AI Guardrails, while Guardrails are also available as a standalone runtime layer.

Check Point's broader AI security portfolio includes Workforce AI Security for employee usage. These are related offerings with different deployment and access requirements, so their features should be evaluated separately.

Features & Capabilities

The Guard API screens interactions for prompt manipulation, content-policy breaches, data exposure, and off-policy agent behavior. A SaaS dashboard provides request investigations, policies, projects, and API-key management. Enterprise options add access controls, log exports, and retention settings.

Red teaming probes applications and agents before deployment. Self-hosted Guardrails provide another deployment option, but the SaaS dashboard is not available to self-hosting customers. Agent discovery and risk assessment are broader capabilities than the standalone Guardrails tier.

User Ratings and Testimonials

Vendor-selected testimonials include Dropbox discussing protection for its LLM-powered applications. These deployment examples are useful context, but do not establish the same performance for every workload. Test false positives, permitted tool actions, and the application's behavior when a request is flagged.

Pricing & Value

Enterprise access uses sales-led commercial terms. Community allows up to 10,000 screening requests per month.

Confirm the chosen tier, request volume, deployment option, and included discovery or testing capabilities. Workforce controls and self-hosted Guardrails should not be assumed to share every SaaS dashboard feature.