AI Guardrails Should Be Rules You Own

Filters you cannot see into, tune, or explain train people to route around them. Guardrails work when the people they affect understand the rule, can change it, and can see what it did.

A guardrail that gets switched off protects nobody.

• Cloud and developer credentials: AWS access keys, GitHub, GitLab, Slack, Stripe, npm, Hugging Face, and SendGrid tokens
• AI provider keys: Anthropic, OpenAI, and Google
• Private keys, bearer tokens, and password or secret assignments
• Cloptima's own tokens and virtual keys, so a leaked key in a prompt is caught too

Treating the two differently is what keeps false positives low. You are not asked to trust one global detector with your national ID formats.

The sides are independent. A common setup redacts credentials in prompts so the work continues, and blocks them in responses so nothing leaks back out.

Before:  Debug this config: api_key = abc123def456ghi789
After:   Debug this config: [REDACTED_SECRET]

resource "cloptima_llm_guardrail_profile" "customer_data" {
  name = "Customer data protection"
  definition = jsonencode({
    input = {
      action    = "redact"
      detectors = { secret = {} }
      custom_rules = [
        { id = "us_ssn", match = { regex = "\\b\\d{3}-\\d{2}-\\d{4}\\b" } },
        { id = "codename", match = { terms = ["Project Falcon"], whole_word = true }, action = "block" },
      ]
    }
    output = {
      action    = "redact"
      detectors = { secret = {} }
    }
  })
}

Rules run on a linear-time matcher. In plain words, that is a guarantee: no pattern, however it is written, can make a request slower or stall other traffic. To keep that promise, rules have sensible bounds: up to 10 terms per rule, up to 64 characters per term or pattern, and no look-ahead or back-references.

Each profile is evaluated as written and the strictest outcome wins. A team profile can add protection, and it cannot weaken the baseline.

1. 1

   Week 1: observe everything

   Create the baseline from the Monitor only template. Review what it finds in the audit log.

2. 2

   Week 2: redact credentials

   Switch the credential rules to Redact. Work continues, and secrets stop reaching models.

3. 3

   Week 3: add personal-data rules in Observe

   Start from the US or European template. Tune rules that match too much.

4. 4

   Week 4: block what must never leave

   Move the rules that matter most to Block, with a clear message for developers.

Local rules run first, so a request they block never pays for a provider scan. You can cap what a provider scan may cost per request, skip it when a request is already slow, and choose what happens if it cannot run. Cloptima never adds a check you did not configure.

• Profile changes are versioned, with who changed what and when
• Credential and custom-rule findings that pass through in Observe appear in the audit log
• A blocked request names the rule that matched, never the content that matched it
• The response a developer sees names the rule, so the fix is obvious