Check Point AI agent security
Check Point documents agent discovery for Amazon Bedrock, Bedrock AgentCore, Google Cloud, Microsoft Copilot Studio and Salesforce Agentforce, recording each agent's owner, model, tools, MCP servers and recent activity in a shared inventory. It says AI Agent Security assesses risk per agent and evaluates tool calls in context, with allow and deny lists and a beta deviation detector. A Guard API screens prompts, tool calls and responses, and AI Guardrails deploy as SaaS, Helm, Docker or air-gapped. Documentation and an OpenAPI spec are public. Agent identity, token and revocation controls: none found on 16 Nov 2026, and native platform runtime integrations are roadmap only.
Facts read on .
Strength and weakness for AI Agent Security Solutions
Strength: Agent discovery across seven named platforms with per-agent risk ratings plus runtime tool-call blocking through one Guard API (check-point-15, check-point-21, check-point-25).
Weakness: No agent identity, token or revocation controls found on 2026-11-16, and native platform runtime integrations are roadmap only (check-point-19, check-point-17).
Scores and evidence
Check Point is 12th in the overall ranking at 56.2 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates.
| Criterion | Score | Weight | Why | Facts |
|---|---|---|---|---|
| Coverage: endpoint layer | 4 of 10 | 6 | An optional desktop agent covers code assistants and desktop agents; OS list, local MCP controls and CLI coverage none found. | check-point-08, check-point-09, check-point-10, check-point-11 |
| Coverage: browser layer | 5 of 10 | 6 | A browser extension redacts prompts and restricts paste for public AI tools; browser agent control none found. | check-point-12, check-point-13, check-point-14 |
| Coverage: cloud workspace layer | 6 of 10 | 6 | Discovery for Bedrock, AgentCore, Copilot Studio and Agentforce is documented; native runtime integrations are roadmap and the product is early access. | check-point-15, check-point-16, check-point-17, check-point-18 |
| Coverage: identity layer | 1 of 10 | 6 | Per-agent risk across tools is assessed, but the docs defer identity to other products; credential inventory and revocation none found. | check-point-19, check-point-20, check-point-21 |
| Coverage: gateway layer | 6 of 10 | 6 | The Guard API screens prompts, tool calls and responses inline; a dedicated gateway is only a planned partnership. | check-point-22, check-point-23, check-point-24 |
| Stops the action | 6 of 10 | 20 | Tool allow and deny lists and a beta deviation detector act at runtime; the pre-execution monitor is research with unstated status. | check-point-25, check-point-26, check-point-27, check-point-28 |
| Visibility | 7 of 10 | 15 | Discovery records each agent's owner, model, tools, MCP servers and recent activity in a shared inventory; vendor documentation only. | check-point-15, check-point-16, check-point-21, check-point-29 |
| Deployment | 6 of 10 | 10 | SaaS, Helm, Docker, air-gapped and extension options are documented; a time to deploy for agent security none found. | check-point-10, check-point-12, check-point-30, check-point-31, check-point-32 |
| Policy and governance | 4 of 10 | 10 | Unified audit trails and a policy simulator exist; SIEM, EDR, IdP and ticketing integrations none found. | check-point-27, check-point-33, check-point-34, check-point-35 |
| Transparency | 7 of 10 | 10 | Documentation, an OpenAPI spec and research are public; a published price none found. | check-point-37, check-point-39, check-point-38 |
| Proof | 7 of 10 | 5 | Third parties report the acquisition and launch; named customers include Dropbox, Nubank and Clalit. | check-point-02, check-point-03, check-point-40, check-point-41 |
Where it ranks
| View | Rank | Total |
|---|---|---|
| Overall ranking | 12th | 56.2 |
| Ranked for the endpoint layer | 13th | 55.0 |
| Ranked for the browser layer | 8th | 58.0 |
| Ranked for the cloud workspace layer | 8th | 61.0 |
| Ranked for the identity layer | 14th | 46.0 |
| Ranked for the gateway layer | 9th | 61.0 |
Product, deployment, price and owner
- Product assessed: Check Point AI Agent Security (discovery, risk assessment, runtime via AI Guardrails) plus Workforce AI Security for endpoint and browser coverage
- Deployment: Managed SaaS or self-hosted (Kubernetes Helm, Docker, air-gapped) for AI Guardrails via API; browser extension plus optional desktop agent for Workforce AI Security; agent platform connectors for discovery
- Time to deploy: Workforce AI Security says it deploys instantly through a browser extension; none found on 2026-11-16 for AI Agent Security and AI Guardrails
- Price: none found on 2026-11-16
- Owner: Check Point Software Technologies Ltd.
- Website: checkpoint.com
Facts and sources
check-point-01 (vendor): Check Point Software Technologies announced on 2025-09-16 that it agreed to acquire Lakera, an AI-native security platform; the press release says Lakera products became Check Point AI Red Teaming and Check Point AI Agent Security.
Quote: "Its flagship solutions, Check Point AI Red Teaming and Check Point AI Agent Security, provide pre-deployment posture assessments and real-time runtime enforcement"
Source: checkpoint.com, read .
check-point-02 (third-party): SecurityWeek reported on 2025-09-16 that Check Point planned to acquire Lakera, and that on closing Lakera would form the basis of a Global Center of Excellence for AI Security.
Quote: "Upon closing, Lakera will form the basis of Check Point’s new Global Center of Excellence for AI Security"
Source: securityweek.com, read .
check-point-03 (third-party): Help Net Security reported on 2026-03-24 that Check Point launched the AI Defense Plane, built on the Lakera and Cyata acquisitions, with workforce, AI application and agent, and red teaming modules.
Quote: "strengthened by technologies from ThreatCloud AI and the company’s recent Lakera and Cyata acquisitions"
Source: helpnetsecurity.com, read .
check-point-04 (vendor): Check Point lists Workforce AI Security, AI Red Teaming and AI Agent Security as its AI Security products, and describes an AI Defense Plane to discover, protect and govern AI.
Quote: "Check Point delivers one platform to discover, protect, and govern every AI interaction from employees to agents."
Source: checkpoint.com, read .
check-point-05 (vendor): The Lakera documentation site is now branded Check Point AI Security, with products AI Agent Security, AI Guardrails and AI Red Teaming.
Quote: "AI Guardrails is the runtime layer of AI Agent Security, and is also sold on its own"
Source: docs.lakera.ai, read .
check-point-06 (vendor): The Lakera website still operates as lakera.ai and describes workforce AI security, AI agent security and AI red teaming products.
Quote: "Shadow AI discovery across apps and browsers"
Source: lakera.ai, read .
check-point-07 (vendor): A 2026-10-01 Check Point press release names Adam Ely as General Manager, AI Security at Check Point.
Quote: "said Adam Ely, General Manager, AI Security at Check Point Software Technologies"
Source: checkpoint.com, read .
check-point-08 (vendor): Check Point says Workforce AI Security covers desktop agents, code assistants such as GitHub Copilot and Cursor, and MCP-connected workflows; the Enterprise tier covers web, desktop, MSP and IDEs.
Quote: "The Enterprise tier covers web, desktop, MSP, and IDEs, while the Essentials tier focuses on web applications."
Source: checkpoint.com, read .
check-point-09 (vendor): Check Point says Workforce AI Security discovers AI tools across web apps, desktop agents, browser extensions, SaaS integrations, code assistants and MCPs, with an application catalog and risk scoring.
Quote: "automatically discovers every AI tool in use, sanctioned or unsanctioned across web apps, desktop agents, browser extensions, SaaS integrations, code assistants, and MCPs"
Source: checkpoint.com, read .
check-point-10 (vendor): Check Point says endpoint coverage comes from an optional desktop agent, while the base deployment is a browser extension.
Quote: "Desktop agent deployment is also available for broader coverage across device-level AI usage."
Source: checkpoint.com, read .
check-point-11 (none-found): None found on 2026-11-16: looked for named operating systems, local MCP server controls and command-line agent coverage for the desktop agent on the Workforce AI Security page and the docs index.
Source: checkpoint.com, read .
check-point-12 (vendor): Check Point says Workforce AI Security deploys through a lightweight browser extension with no architecture changes.
Quote: "Workforce AI Security deploys instantly through a lightweight browser extension, no complex setup, no downtime, and no architecture changes required."
Source: checkpoint.com, read .
check-point-13 (vendor): Check Point says Workforce AI Security protects data in prompts with inline redaction, copy and paste restrictions, file and image redaction with OCR, and user guidance.
Quote: "Protections include inline redaction, copy/paste restrictions, file and image redaction with OCR, and interactive user guidance."
Source: checkpoint.com, read .
check-point-14 (vendor): Check Point says Workforce AI Security covers public AI tools, browser extensions and SaaS-integrated AI features.
Quote: "Workforce AI Security covers the full spectrum of employee AI interactions: public AI tools (ChatGPT, Gemini, Claude, Copilot), browser extensions, desktop agents"
Source: checkpoint.com, read .
check-point-15 (vendor): Check Point documents agent discovery for Amazon Bedrock, Bedrock AgentCore, Google Cloud, Microsoft Copilot Studio, Salesforce Agentforce, n8n and Relevance AI.
Quote: "Discovery is available today for: Amazon Bedrock Amazon Bedrock AgentCore Google Cloud Microsoft Copilot Studio Salesforce Agentforce n8n Relevance AI"
Source: docs.lakera.ai, read .
check-point-16 (vendor): Check Point documents that discovery records each agent's name, owner, model, tools, connected MCP servers and recent activity, with depth varying by platform.
Quote: "Connected MCP servers and the tools they provide."
Source: docs.lakera.ai, read .
check-point-17 (vendor): Check Point documents that native platform runtime integrations are on the roadmap, so runtime protection for cloud-hosted agents goes through the Guard API today.
Quote: "native platform runtime integrations are on the roadmap"
Source: docs.lakera.ai, read .
check-point-18 (vendor): Check Point states that AI Agent Security entered Early Access on 2026-06-01, adding discovery and risk assessment to AI Guardrails.
Quote: "Check Point AI Agent Security extends AI Guardrails with discovery and risk assessment for the agents your organization builds and deploys"
Source: docs.lakera.ai, read .
check-point-19 (none-found): None found on 2026-11-16: looked for agent identity inventory, token or credential management and permission revocation on the AI Agent Security page and docs; the docs defer identity to other products.
Source: docs.lakera.ai, read .
check-point-20 (vendor): Check Point docs say identity, cloud and network controls sit outside AI Agent Security and come from other parts of the architecture.
Quote: "controls elsewhere in your architecture, such as your identity provider, cloud security, and network controls"
Source: docs.lakera.ai, read .
check-point-21 (vendor): Check Point says AI Agent Security assesses agent risk across tools, skills, MCP servers, usage and connected components, and the docs add per-agent risk ratings.
Quote: "Understand agent risk across tools, skills, MCP servers, usage, and connected components."
Source: checkpoint.com, read .
check-point-22 (vendor): Check Point documents the Guard API as the single integration point that screens prompts, model outputs, tool calls, tool responses and tool descriptions for an agent workflow.
Quote: "applied across the agentic workflow via the Guard API: user prompts, model outputs, tool calls, tool responses, and tool descriptions"
Source: docs.lakera.ai, read .
check-point-23 (vendor): Check Point says AI Agent Security acts as an inline enforcement layer that inspects prompts, responses, external content and agent tool calls before risky actions run.
Quote: "AI Agent Security acts as an inline enforcement layer for AI agent interactions."
Source: checkpoint.com, read .
check-point-24 (vendor): On 2026-10-01 Check Point and Clalit Health Services announced a partnership to jointly build a dedicated AI Gateway between users, applications and AI models.
Quote: "Check Point and Clalit development teams will jointly build a dedicated control and security layer between users, applications, and AI models."
Source: checkpoint.com, read .
check-point-25 (vendor): Check Point says AI Agent Security evaluates agent tool calls in context to help block unsafe or unauthorized behavior at runtime.
Quote: "Evaluate agent tool calls and actions in context, helping block unsafe or unauthorized behavior at runtime."
Source: checkpoint.com, read .
check-point-26 (vendor): Check Point docs say the Dangerous Deviation detector (beta) flags agent tool calls that cause data leakage, privilege escalation or system destruction with no warrant in the trusted conversation.
Quote: "Access and privilege escalation : public sharing of resources, sudo or admin grants, widened ACLs, or attempts to escalate permissions beyond a sandbox."
Source: docs.lakera.ai, read .
check-point-27 (vendor): Check Point docs describe a Tool Allow/Deny List enforced through Guard API screening, and a project Detect mode that monitors without blocking.
Quote: "whether the tools it calls are allowed at all"
Source: docs.lakera.ai, read .
check-point-28 (vendor): A Check Point research blog dated 2026-09-15 describes a sidecar control monitor that prevents harmful agent actions before they execute at sub-100ms latency; the post is research, and its product status is not stated there.
Quote: "prevents harmful actions before they execute in real time at sub-100ms latency"
Source: blog.checkpoint.com, read .
check-point-29 (vendor): Check Point says the agent inventory is shared between AI Agent Security and Workforce AI Security, so on-device agents appear alongside platform agents.
Quote: "agents discovered on employee devices appear through Workforce AI Security"
Source: docs.lakera.ai, read .
check-point-30 (vendor): Check Point documents AI Guardrails as managed SaaS or self-hosted on Kubernetes (Helm), Docker, or air-gapped.
Quote: "AI Guardrails is available as a fully managed SaaS solution or as a self-hosted deployment ."
Source: docs.lakera.ai, read .
check-point-31 (vendor): Check Point documents that runtime protection integrates through the Guard API, with discovery and risk assessment managed in the portal.
Quote: "Discovery and risk assessment are posture capabilities, managed in the portal."
Source: docs.lakera.ai, read .
check-point-32 (none-found): None found on 2026-11-16: looked for a published time to deploy for AI Agent Security and AI Guardrails in the docs quickstart and product pages; only Workforce AI claims "instantly".
Source: docs.lakera.ai, read .
check-point-33 (vendor): Check Point says the AI Defense Plane applies consistent policy with unified audit trails and compliance reporting.
Quote: "Apply consistent policy across all AI interactions, with unified audit trails and compliance reporting."
Source: checkpoint.com, read .
check-point-34 (vendor): Check Point added a Policy Impact Simulator on 2026-04-10 that shows how sensitivity levels would have affected historical traffic.
Quote: "an interactive tool that shows how different sensitivity levels and guardrail configurations would have affected your historical traffic"
Source: docs.lakera.ai, read .
check-point-35 (none-found): None found on 2026-11-16: looked for named SIEM, EDR, identity provider and ticketing integrations for AI Agent Security and AI Guardrails on the product pages, integration guide and docs index.
Source: docs.lakera.ai, read .
check-point-36 (vendor): Check Point docs provide a request latency percentiles chart on the Analytics page (2026-10-26) and a Guard logs API.
Quote: "Request latency percentiles chart on the Analytics page"
Source: docs.lakera.ai, read .
check-point-37 (vendor): Check Point publishes AI Security documentation, an OpenAPI spec and a changelog that are readable without sign-in.
Quote: "Subscribe via RSS"
Source: docs.lakera.ai, read .
check-point-38 (none-found): None found on 2026-11-16: looked for a published price for AI Agent Security, Workforce AI Security and AI Guardrails on the Check Point AI Security pages and the Lakera site; the pages only offer demos.
Source: checkpoint.com, read .
check-point-39 (vendor): Check Point publishes AI agent security research on its blog, including a post dated 2026-09-15 on synchronous control monitoring.
Quote: "Control monitors ingest the trace as context."
Source: blog.checkpoint.com, read .
check-point-40 (vendor): The AI Agent Security page quotes a Dropbox security engineer and the Nubank CISO as customers of the Lakera technology.
Quote: "The Lakera team has accelerated our GenAI journey, allowing us to create secure GenAI experiences at scale."
Source: checkpoint.com, read .
check-point-41 (vendor): On 2026-10-01 Check Point announced Clalit Health Services will deploy Check Point AI Security for real-time visibility and control over AI usage and agents.
Quote: "Check Point AI Security will give Clalit real-time visibility and control over AI usage and AI agents across the organization."
Source: checkpoint.com, read .
check-point-52 (vendor): On 2026-07-30 Check Point announced a new AI Network Firewall, listed in its newsroom.
The exact quote is on the source page.
Source: checkpoint.com, read .