Security controls
Runtime controls for di research environment — di rules wey dey show here na di same rules wey di side wey dey allow request dey actually check.
Guardrails
kotoba-guardrails-2026-09-v2Di content policy wey dey apply to di body of every request. Decisions dey deterministic; if e block, e go refuse di job before any quota spend.
| Name | Action | Description | Patterns | E dey active |
|---|---|---|---|---|
csam-block
|
Block | Dem dey refuse CSAM references for every request (AUP strictly forbid am). | 0 | Always on (AUP) |
cbrn-block
|
Block | Dem dey refuse CBRN/WMD uplift request for every request (AUP strictly forbid am). | 0 | Always on (AUP) |
fraud-block
|
Block | Dem dey refuse any fraud-as-a-service request (AUP strictly forbid am). | 0 | Always on (AUP) |
secret-hygiene
|
Cover | Any live credential shape wey dey di prompt, dem go mask am before di model see am. | 3 | |
pii-passkeys
|
Flag | Any mention of passkeys/private keys, dem go flag am for review, dem no go block am at all. | 2 |
Firewall
kotoba-firewall-2026-09-v1Di task tools wey agent fit use and di trust rung wey each one need. If call fall below im rung, dem go refuse am (observe mode go just record am).
| Tool | Level wey dem need | Tasks wey still dey open |
|---|---|---|
code-review
|
Identity don verify | code-review |
vulnerability-triage
|
Identity don verify | vulnerability-triage |
remediation
|
Identity don verify | remediation |
payload-crafting
|
Contract don sign | — (dem deny am by default) |
c2-tooling
|
Contract don sign | — (dem deny am by default) |
Compliance
kotoba-compliance-2026-09-v1Automated runtime controls mapped to each framework — a description of controls, not a certification, audit or compliance attestation. No certification is held (see /trust/). PCI DSS scope is delegated to Stripe: card data never reaches Mithril.
| Controls mapped to (not a certification) | Region | Automatic controls | E dey active |
|---|---|---|---|
| Automated controls mapped to APPI | JP | prompt-pii-redaction audit-log-retention screening-evidence-receipts | |
| Automated controls mapped to PCI DSS 4.0 Delegated to Stripe — card data never reaches Mithril |
Global | card-data-never-stored secret-hygiene-redaction audit-log-retention | |
| Automated controls mapped to OWASP LLM Top-10 | Global | guardrails-csam-cbrn firewall-task-deny screen-model-block-receipts | |
| Automated controls mapped to NIST AI RMF | US | session-projection-uncalibrated assurance-ladder-evidence |
How long audit log go stay
180 days
Where data dey stay
Not pinned — no storage or processing region is selected in code or configuration (Cloudflare's global network and providers whose regions are not specified; see /legal/subprocessors/)
No certification is held. Trust Center
Data retention
No inference logs / no training (conversation text kept until account delete)