Aipedia: Company formation and secretarial agents
Corporate service providers should keep unpublished cap tables, residential addresses and identity documents out of public AI tools. A leak can breach confidentiality and UK GDPR, and the resulting costs may fall between PI and cyber cover.
Part of: AI in Company formation and secretarial agents
In short
Formation agents and company secretaries hold sensitive information before it becomes public: share structures, shareholder agreements, PSC details and officers' residential addresses, which are protected information. Identity verification adds passports and biometric selfie data, which is special category data when used to identify someone. Putting any of this into a public AI tool risks a confidentiality breach, and AI identity tools need a lawful basis and a data protection impact assessment. PI may respond to a client's confidentiality claim, subject to the wording, while breach response costs usually sit with cyber insurance.
Corporate service providers see information long before it reaches the register, and some of it never does. The main categories are these:
AI tools are used across all of this: summarising agreements, drafting minutes, analysing structures and running identity checks. Each use is a decision about where the data goes.
For the sector overview, see AI in company formations.
Pasting a client's cap table or shareholder agreement into a public generative AI chatbot can disclose it to a provider whose terms may allow inputs to be retained or used. That can breach the agent's duty of confidence and its engagement terms, even if nothing is ever published. The safest rule is simple: these categories of information do not go into public AI tools.
Firms that want the benefits of AI should use tools with contractual protections on confidentiality, data location and retention, and should tell clients how their information is processed.
Identity verification brings in the most sensitive data. Where biometric data is used to identify someone, it is special category data under UK GDPR. An AI identity tool therefore needs a lawful basis and a condition for special category processing, and because the processing is high risk, a data protection impact assessment (DPIA) is needed before it goes live. The ICO's guidance on AI and data protection covers fairness, transparency and accuracy for AI systems.
Automated outcomes raise a further point. An automated client-acceptance or verification refusal may be a solely automated significant decision under UK GDPR Articles 22A to 22D, in force from 5 February 2026 under section 80 of the Data (Use and Access) Act 2025. Safeguards, including a route to human review, are then required. For the consultancy view, see advising on facial recognition and biometrics and advising on DPIAs for AI systems.
Illustrative scenario: a company secretary uploads a client's draft shareholder agreement, cap table and investor list to a free AI tool to produce a summary for the board. The tool's terms allow inputs to be retained. Separately, the firm's AI identity verification provider suffers a breach exposing selfie images and passport scans of several clients' directors. The firm has to assess whether to report to the ICO, notify the people affected and respond to clients who say their confidential deal information was mishandled.
A client's claim that the agent breached confidentiality and caused it loss may fall within PI, subject to the wording; many PI policies include breach of confidence or data protection liability, though some limit it. The first-party costs of a breach, such as forensic investigation, notifying individuals, legal advice on reporting and restoring systems, are usually covered by cyber insurance rather than PI. ICO fines are commonly excluded or uninsurable. A breach at a vendor may still leave the agent answerable to its clients. See GDPR and data breach cover within PI and combining cyber and PI.
These steps keep client data under control when AI is involved:
Apex arranges PI and cyber cover for corporate service providers. A broker maps how your policies split confidentiality claims and breach response costs for AI tools and vendors.
With care. Unpublished agreements and cap tables are confidential, so they should not go into public AI tools. A tool with contractual protections on retention and data use, together with client agreement where needed, is a safer route.
Yes, in practice. Face matching uses biometric data to identify people, which is special category data, and the processing is high risk. A DPIA should be completed before the tool is used, alongside a lawful basis and a special category condition.
Partly, at most. A client's claim for breach of confidence may fall within PI, subject to the wording. Investigation, notification and recovery costs are usually cyber insurance matters, and ICO fines are commonly excluded or uninsurable.
Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.
Start a PI quote Request a callbackApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.