Use: complete when a Neuroscale AI agent (an AI system that conducts conversations or takes actions, e.g., outreach or scheduling agents) is configured for production, and on material change. Owner: product owner + CISO. Links to the model card. Frameworks: ISO/IEC 42001 A.6.2 / A.9.4; EU AI Act Art. 50; CA SB 1001 (§17941) bot disclosure; supports outbound-communications controls.
An “agent” here is an AI system that interacts with a person or acts on behalf of a Neuroscale service. Replace every {{var}}.
1. Agent identification
- Agent name / purpose: {{agent_name_purpose}}
- Underlying model & version: {{model_and_version}} (link the model card)
- Channels: {{channels}}
- Acting on behalf of: {{acting_on_behalf_of}}
2. Disclosure verification
- AI-interaction disclosure shown: {{disclosure_shown}} per the AI Interaction Disclosure (EU AI Act Art. 50; CA SB 1001 §17941).
- Disclosure cadence: {{disclosure_cadence}}
- Human hand-off path: {{human_handoff_path}}
3. Authorization & boundaries
- Authorized actions: {{authorized_actions}}
- Prohibited actions / guardrails: {{prohibited_actions}}
- Identity & authentication of the agent to systems: {{agent_authentication}}
- Rate / scope limits: {{rate_scope_limits}}
4. Compliance checks
| Check | Status |
|---|---|
| AI-interaction disclosure verified | {{check_disclosure_verified}} |
| Outbound consent / suppression honored (if outreach) | {{check_consent_honored}} — see Outbound Communications Policy |
| Logging of agent interactions enabled | {{check_logging_enabled}} |
| Human oversight / kill-switch in place | {{check_human_oversight}} |
| Deployment approved | {{deployment_approval_link}} (link AI Deployment Approval Record) |
5. Sign-off
- Verified by: {{verified_by}} — {{verification_date}}
- Re-verification date: {{re_verification_date}}
Variables
| Variable | Description |
|---|---|
{{agent_name_purpose}} | Agent name and purpose (e.g., candidate-outreach agent) |
{{model_and_version}} | Underlying model and version |
{{channels}} | Channels the agent operates on (email / SMS / voice / in-app chat) |
{{acting_on_behalf_of}} | Whether the agent acts for Neuroscale or a Customer (deployer) |
{{disclosure_shown}} | Whether the AI-interaction disclosure is shown (Y/N) and its exact text/placement |
{{disclosure_cadence}} | Disclosure cadence (first message / per session / per channel rules) |
{{human_handoff_path}} | How a person reaches a human |
{{authorized_actions}} | What the agent may do |
{{prohibited_actions}} | What the agent must not do; escalation triggers |
{{agent_authentication}} | Service credential / scope; least privilege per Access Control Policy |
{{rate_scope_limits}} | Rate / scope limits |
{{check_disclosure_verified}} | AI-interaction disclosure verified (Y/N) |
{{check_consent_honored}} | Outbound consent / suppression honored (Y/N) |
{{check_logging_enabled}} | Logging of agent interactions enabled (Y/N) |
{{check_human_oversight}} | Human oversight / kill-switch in place (Y/N) |
{{deployment_approval_link}} | Link to the AI Deployment Approval Record |
{{verified_by}} | Verifier(s) — product owner; CISO |
{{verification_date}} | Date of verification |
{{re_verification_date}} | Re-verification date (default 24 months; sooner on material change) |