Answers

The questions regulated teams actually ask

Direct answers first, depth after. Every page opens with a ~100-word answer, cites primary sources, and shows when it was last reviewed — because most of these questions now get asked to an AI assistant, and the assistant deserves a source worth citing.

Legal AI

Can law firms use ChatGPT with client data?

Not with identifiable client data in the clear. The Law Society’s guidance says confidential data should not go into free, public generative AI tools, and the Upper Tribunal accepted in Munir v SSHD that pasting client correspondence into ChatGPT was a data breach. Law firms can still use ChatGPT productively in two ways: under an enterprise contract with retention and no-training guarantees, or — for any tool — by masking client-identifiable information before the prompt leaves the firm, so the model receives a de-identified version and confidentiality is never in question.

Read the answer

Practical controls

How do I stop staff pasting PII into ChatGPT?

You cannot reliably stop it with policy alone — you make it technically impossible for the PII to reach the tool. The four options are: ban AI (usage moves to personal devices and you lose visibility), train staff (helps, but fails under deadline pressure), buy an enterprise AI contract (covers one tool only), or mask PII at the point of use, in the browser, before the prompt leaves. Masking is the only option that works regardless of which AI tool staff choose, without slowing them down — and it produces an audit record instead of a blind spot.

Read the answer

Technology

What is reversible PII tokenization?

Reversible PII tokenization replaces sensitive values in a prompt with consistent placeholders — "Sarah Thompson" becomes <PERSON_7K9X> — before the text reaches an AI model, then restores the real values in the response after it returns. Unlike permanent redaction, the answer comes back readable; unlike blocking, the work continues. The mapping between placeholder and real value lives in an encrypted vault (AES-256-GCM) with a short time-to-live — 15 minutes by default — so the restore window closes automatically. The AI provider only ever sees placeholders.

Read the answer

What is DLP for LLMs?

DLP (data loss prevention) for LLMs is the control layer that stops sensitive data leaving your organisation through AI prompts and uploads. Traditional DLP watches files at rest, email, and endpoints — but text typed into ChatGPT’s input box usually walks straight past it. LLM-focused DLP closes that gap with three possible responses: block the prompt (stops work), flag it for review (after the fact), or mask the sensitive values and forward the rest (work continues, data stays). The architectures differ mainly in where they sit — network edge, endpoint agent, or in the browser at the point of use.

Read the answer

Case law

What did the UK tribunal say about ChatGPT and client data?

In Munir v SSHD [2026] UKUT 81 (IAC) — promulgated in November 2025 — the Upper Tribunal dealt with a legal representative who had used ChatGPT to summarise Home Office decision letters and improve client emails. Asked to explain his process at a hearing, he accepted it amounted to a data breach and undertook to notify his clients and regulators. The tribunal characterised uploading such material to an open AI tool as placing it in the public domain, breaching client confidentiality — and distinguished open tools from closed systems running inside a secure network.

Read the answer

Risk

What is shadow AI?

Shadow AI is employees using AI tools — ChatGPT, Claude, Gemini, translation and summarisation apps — outside any sanctioned, visible channel. It is the AI-era version of shadow IT, with a sharper edge: the input is often confidential text pasted in by the very people producing the most work. Bans reliably make it worse by pushing usage onto personal devices where no control can see it. The practical response is to measure it (discover what is actually being used), then make the sanctioned path safer than the shadow path with a control that masks sensitive data at the point of use.

Read the answer

Data protection

Is ChatGPT GDPR compliant?

The question is slightly wrong — and the distinction matters. GDPR compliance is a property of processing, not of a tool: the ICO is explicit that accountability lies with the controller, i.e. the organisation using it. What you can assess is OpenAI’s data terms. Consumer ChatGPT trains on your conversations by default unless you opt out. Business products are different: ChatGPT Business, Enterprise, and the API do not train on your data by default, support a Data Processing Addendum, and offer retention controls. Whether your use is compliant depends on what personal data you put in, on what lawful basis — which is where minimisation and masking come in.

Read the answer

Regulation

Does the EU AI Act apply to UK firms?

Usually not — merely using ChatGPT in a UK office does not bring a UK firm under the EU AI Act. The Act (Regulation (EU) 2024/1689, in force since August 2024) reaches beyond the EU in two main ways: if you place an AI system on the EU market as a provider, or if the output of your AI use is used in the EU. A UK firm serving EU clients with AI-produced work can be caught by that second trigger. What binds UK firms day to day is UK law: DUAA-amended UK GDPR and sector regulators, not the AI Act.

Read the answer

About these pages

Who writes these answers?

The NeutralAI team — an AI security engineering group focused on PII detection and masking for regulated UK teams. Every date, quote, and statistic is checked against primary sources before publication, and each page shows when it was last reviewed.

Are these pages legal advice?

No. They summarise guidance, law, and technology honestly for orientation. For decisions about your firm’s specific obligations, take advice from your own counsel or DPO — and read the primary sources we link on every page.

Looking for the regulator detail?

The UK compliance hub maps Law Society, SRA, Bar Council, judiciary, ICO, DUAA, and FCA guidance to concrete technical controls — with verbatim lines and source links.

Open the compliance hub