Rules
A Firewall for AI
Cerbera AI works like a firewall. There is a catalog of standard rules that reflects what is common across most organizations, and you can write your own. Each rule defines what it targets and what action to take when it matches.
Because the rules are signature-based rather than predictive, they are fast and run reliably on any device, including older fleets. See How It Works for the performance characteristics.
The Four Actions
| Action | What happens | Effect on the user |
|---|---|---|
| Allow | The request is explicitly permitted, overriding a broader block. | None. Silent. |
| Monitor | The match is logged. Nothing is changed or stopped. | None. Silent. |
| Redact | Sensitive data is stripped from the request before it reaches the provider. | Usually none. The workflow continues. |
| Block | The request is stopped. | A pop-up explains what was intercepted and why. |
Allow
Explicitly permits traffic that a broader rule (or your default policy) would otherwise catch. Use it to carve out an exception, for example a specific AI provider or MCP server that should never be blocked regardless of other rules.
Monitor
The starting point for every deployment. Rules log matches without changing anything, so you can learn usage and tune policy with zero user impact.
Redact
When a request contains secrets such as API keys, AWS keys, tokens, or PII, Cerbera redacts them before the request leaves the device, so they are never sent to Anthropic, OpenAI, or any other provider. Redaction rarely interrupts the workflow, because for most workflows sending the key to the provider was not actually necessary. The benefit is that you avoid exposing credentials you would then have to rotate.
Block
The request is stopped and the user sees a pop-up explaining what happened at the moment the request was intercepted. Blocking is how you enforce hard limits, for example disallowing a specific provider or a risky agent behavior.
Building a Rule
Templates
Most rules start from a template, so you do not have to know the exact API of each tool. Templates are grouped by what they detect: common secrets (API keys, cloud credentials, tokens), PII, and AI provider traffic (ChatGPT, Claude, and others). Pick one from the template browser and adjust it rather than writing it from scratch.
AI-assisted composer
Describe the rule you want in plain language and Cerbera drafts it for you: the target, the match conditions, and the action. Review and adjust before saving.
Custom rules
Write a fully custom rule. A rule can match any domain, so you could, for example, block an arbitrary site the same way you block an AI tool.
What a Rule Can Match
A rule can be scoped along several dimensions:
- Target: apply to all AI providers, or scope to specific ones.
- Who it applies to: all users, or a specific set of user/device groups.
- Content: the secret or PII templates (and any custom match conditions) it should detect.
- Account posture: license checks that flag personal (non-enterprise) account usage on providers like Claude and ChatGPT.
- Agent posture: a check for risky agent configuration, for example a coding agent like Codex CLI running with approvals turned off ("YOLO mode").
- Which action to take: allow, monitor, redact, or block.
Simulating a Rule
Before you save a rule, or when you want to sanity-check an existing one, the rule simulator lets you run it against sample or historical traffic and see what it would have matched, without affecting live enforcement.
Draft the rule
Start from a template, the AI composer, or a blank rule.
Simulate it
Run the simulator to see which events it would have matched.
Check performance
The Rules page includes a performance view showing rule evaluation latency, so you can confirm a new rule is not adding meaningful overhead before you ship it.
Promote it
Start in monitor, confirm it matches what you expect in production, then promote it to redact or block.
How Rules Reach the Device
You define rules in the Cerbera cloud. The local agent authenticates to Cerbera over HTTPS, pulls the current rule set, and caches it on disk so enforcement keeps working even if the device goes offline or Cerbera is briefly unreachable. On the next successful check, only rules that changed are re-downloaded.
The Managed Catalog
Maintaining rules for a fast-moving ecosystem is exactly the work Cerbera takes on for you. We keep the catalog current, propose new rules when a new tool or risky pattern appears, and tell you what we found and what we recommend. You approve, we implement. The goal is that you fine-tune for your specifics rather than maintain the base from scratch.
Start every rule in monitor, confirm it matches what you expect (using the simulator first), then promote it to redact or block. This mirrors a safe firewall rollout and avoids surprising users.
Export to Your SIEM
Every rule match is also available in OpenTelemetry format, so alerts and logs can flow into your SIEM and the tools where they make the most sense, instead of living only in the Cerbera app. See OpenTelemetry Export for how that export works.