[Firm name]: AI tool and risk register
Owner: [name, role]. Last reviewed: [date]. Next review: [date].
Columns
- Tool: the AI tool, for example ChatGPT.
- Plan: the plan and account type, for example ChatGPT Business, signed in with work accounts.
- Contract: is there a contract and data processing agreement with the firm? Yes or no.
- Data allowed: what may go in, for example "client work, with client details kept out where not needed", or "no personal data".
- Owner: who's responsible for this tool.
- Review date: when the row will next be checked.
- Risk: the main risk, and whether it's low, medium or high.
- Mitigation: what you do about it.
Example rows (made-up example)
Made-up example, not a real firm.
- ChatGPT | Business, work accounts | Yes | Client work, client details kept out where not needed | Office manager | [date] | Medium: client data in prompts | Business plan only; AI policy; client details swapped out before sending.
- ChatGPT | Free or Plus, personal accounts | No | No personal data | Office manager | [date] | High: client data on an account with no contract | Personal accounts warned about or blocked; AI policy.
- Microsoft Copilot | Work account (Entra ID) | Yes | Client work | IT provider | [date] | Medium: web search queries fall under Microsoft's services agreement | Staff told not to put client details into questions that search the web.
- New AI note-taker | Personal sign-up | No | Not approved | Partner | [date] | High: recordings of client meetings stored by an unknown provider | Under review; not to be used in client meetings until approved.
