FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Aipedia: IT consultants and developers
The UK has three main voluntary security codes for IT firms: the AI Cyber Security Code, the Software Security Code and the NCSC secure AI development guidelines. Because IT consultants have no statutory regulator, these documents are among the clearest published statements of what competent practice looks like.
Part of: AI in IT consultants and developers
In short
DSIT's AI Cyber Security Code of Practice (31 January 2025) sets 13 voluntary principles, most aimed at developers and system operators. The Software Security Code of Practice (7 May 2025, updated 15 January 2026) sets 14 principles for software vendors. The NCSC's 2023 guidelines cover secure design, development, deployment and operation of AI. None is legally binding, but in a negligence claim they may be used as evidence of what a reasonably competent developer would have done, which is the test that decides whether a technology PI claim succeeds.
Three government documents speak directly to how IT firms should build and run AI and software securely. A fourth, the BCS Code of Conduct, applies to BCS members and governs professional behaviour generally.
| Document | Publisher and date | Status | Key points for IT firms |
|---|---|---|---|
| AI Cyber Security Code of Practice | DSIT, 31 January 2025 | Voluntary | 13 principles across design, development, deployment, maintenance and end of life; aimed mainly at developers and system operators |
| Software Security Code of Practice | DSIT and NCSC, 7 May 2025, updated 15 January 2026 | Voluntary | 14 principles software vendors are expected to implement, aimed at supply chain attacks and resilience |
| Guidelines for secure AI system development | NCSC, 27 November 2023 | Guidance | Secure design, development, deployment, and operation and maintenance |
| BCS Code of Conduct | BCS | Binding on BCS members | No AI-specific wording; general duties of competence, care, confidentiality and honesty |
The AI Cyber Security Code is being submitted to ETSI as the basis for a technical specification, ETSI TS 104 223, with an implementation guide, TR 104 128. That would give the UK voluntary code a European standards body specification, which may make it more likely to be cited in contracts and tenders.
The Code of Practice for the Cyber Security of AI describes itself as voluntary. It groups duties by lifecycle stage and names the stakeholders each principle mainly applies to. Developers are those "responsible for creating or adapting an AI model and/or system". System operators have "responsibility for embedding / deploying an AI model and system within their infrastructure". Data custodians control data permissions and integrity. An IT consultancy can be more than one of these on a single project.
The principles most relevant to day-to-day consulting work are:
Principle 8 is the one developers most often overlook. If a client later alleges your AI system behaved badly, documentation of training data, model versions and system prompts is often the evidence that shows what you did and why. A plain-English summary by Lewis Silkin sets out the code's minimum expectations.
The Software Security Code of Practice is aimed at software vendors and their customers and is intended to reduce the risk of software supply chain attacks and other resilience incidents. It is not AI-specific, but it applies to any software you sell or maintain, including software written with AI assistance or software that embeds a model. The January 2026 update added details of a Software Security Ambassadors Scheme supporting the code.
For developers who use AI coding assistants, the practical point is simple: the code does not lower expectations because a machine wrote the first draft. Secure development, dependency management and vulnerability handling still apply to every line you ship.
The NCSC's Guidelines for secure AI system development (27 November 2023) are organised around four areas: secure design, secure development, secure deployment, and secure operation and maintenance. They predate the DSIT code and cover similar ground, so many firms map their controls to both. Teams building AI chatbots or AI agents with system access will find the deployment and operation sections particularly relevant.
The split matters when something goes wrong. If you build a model or AI system and hand it to the client to run, you are acting mainly as a developer: documentation, testing and secure design sit with you. If you also host, deploy or manage it, you are acting as a system operator too, taking on monitoring, updates and communication with end users. Many disputes will turn on who agreed to do what, so the role you take should be written into the statement of work. Our guide to AI clauses in IT contracts covers how to do that.
A professional negligence claim against an IT consultant asks whether you used the skill and care of a reasonably competent member of your profession. With no statutory regulator and no AI wording in the BCS Code, published government codes are an obvious reference point for experts on both sides. Failing to follow a widely known code is not automatically negligent, but it may be hard to explain why you did not.
For PI, that means the codes can shape whether a claim succeeds and how it is valued. Technology PI usually covers negligent acts, errors and omissions in your professional services, subject to the policy terms. Showing that your AI work followed the codes may help defend a claim. Some insurers may also ask about AI security practices at renewal, so a short note of which principles you follow can be useful when you complete a proposal form. See what to document about AI use for PI and our overview of AI in IT consulting.
Apex arranges technology PI for IT consultants and developers. A broker can review how your policy treats AI-assisted and AI-built work and whether its terms match the security commitments in your client contracts.
No. The code describes itself as voluntary. DSIT and the NCSC are submitting it to ETSI as the basis for TS 104 223, which may increase its use in contracts and tenders, but there is no legal duty to comply with it in the UK.
Most principles apply to both developers and system operators. Principle 8, documenting data, models and prompts, applies primarily to developers. Principles 1, 5 and 7 also extend to data custodians. Principle 10 names stakeholders provision by provision.
The code is not AI-specific. It applies to software vendors generally, so software written with AI help or software that embeds a model is still within its expectations on secure development and supply chain resilience.
It can. A court asks what a reasonably competent IT professional would have done, and published government codes are likely reference points for expert evidence. Following them will not guarantee a defence but may help show reasonable care.
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.
