Research and drafting
Advisers may use AI to compare published regulatory circulars, check bilingual terminology or draft training scenarios. Before citing a regulatory requirement, they must check the regulator’s original text, who it applies to and when it takes effect.
Client information and approved tools
Public AI tools may be used only with public information or fictional examples. Before processing client files, identity documents or confidential business information, we must confirm the scope of the engagement and the permissions required. We assess access controls, retention and use of data for model training, approve the tool and enter only the information needed.
PCPD checklist for employees’ use of generative AIProfessional judgement and sign-off
The responsible adviser checks AI-assisted work against the evidence in the client file. An authorised person makes the final decision on client risk ratings, independent review conclusions and documents submitted to a regulator.
Errors and follow-up
Anyone who identifies an inaccurate citation, biased output or unintended disclosure must stop using the affected material and notify the responsible adviser. We check which documents are affected, notify clients and issue corrections where needed, and record the action taken.