The safest private ChatGPT alternatives for a UK law firm are business AI accounts that don't train on client data, such as Microsoft 365 Copilot if you're on Microsoft 365, or ChatGPT Enterprise if you need UK storage at rest. If client data must be processed in the UK as well, use your own cloud tenant or on-premise models.
Here are nine options compared on training, retention and data location. Number 1 is our own service, so read that entry accordingly. Last updated 19 November 2026.
Any business plan is a big step up from personal logins, and none of the business plans here train on your data by default.
For UK storage at rest, look at ChatGPT Enterprise or your own cloud tenant. If contracts require UK processing as well, use your own cloud tenant in London or an on-premise model.
What makes a ChatGPT alternative private?
Four requirements matter most, and they match what the SRA's August 2026 warning notice asks about:
- No training on client data. The SRA expects client data not to be used to train AI models unless that's explicitly authorised.
- Clear retention terms. Data is kept no longer than necessary, and you can state for how long.
- Known storage and processing locations. UK GDPR doesn't require UK hosting, though transfers abroad need safeguards and some clients write UK hosting into their contracts. Storage and processing are often in different regions.
- A firm-managed account. The firm controls access and can audit what staff put in, instead of relying on personal logins.
The SRA notes that free and paid tools can both fall short depending on their terms and settings. I've gone tool by tool on the GDPR side in Is ChatGPT GDPR compliant?, and our private AI guide sets out the five levels from a personal login to your own servers.
Private ChatGPT alternatives compared
| Compared on | Trains on your data? | Data stored in the UK? | Price published? |
|---|---|---|---|
| NQU's AI operating system | No | Yes, if you choose UK | No |
| Microsoft 365 Copilot | No | With your Microsoft 365 data. UK processing expected by end of 2026 | Yes, £16.10 a user a month |
| ChatGPT Enterprise | No | Yes, at rest, for new workspaces | No |
| ChatGPT Business | No | No | Yes, $20 a user a month |
| Claude Team and Enterprise | No | No, US only | Yes, from $20 a seat a month |
| Harvey | No | No. EU, Switzerland, US or Australia | No |
| Legora | No | Not publicly. Mainly Sweden | No |
| Lexis+ with Protégé | No | Processed in Europe including the UK. Ask which | No |
| Your own cloud tenant | No | Yes, e.g. Claude on AWS Bedrock London | Pay per use |
| On-premise AI | No | Yes, on your own servers | Hardware cost |
Prices exclude VAT. All details were verified on 19 November 2026, with vendor sources linked in each section below.
1. NQU's private AI operating system (us)
Full disclosure: this one is ours. We build AI into the accounts your firm already runs, usually Microsoft 365 and your practice management system, handling multi-step workflows like intake, filing emails to the right matter and producing first drafts from your precedents. A fee earner reviews and approves client-facing drafts. Everything runs on business AI accounts or your own cloud tenant (levels 2 to 4 on our private AI page), under the data terms of whichever model you choose. More in the AI operating system for law firms.
- Best for
- 10 to 200 person firms that want AI running inside their existing systems rather than in a standalone chat window
- Trains on your data?
- No. Business AI accounts only, with no training on your data.
- Where data lives
- In your own Microsoft or Google account, or ours in the UK, EU or US.
- Retention
- We document data flows as we build, giving your data protection lead the detail needed for a DPIA.
- Pricing
- Depends on firm size. We go through it on the intro call.
- Watch out for
- This is a custom build rolled out one team at a time rather than an off-the-shelf licence, so it isn't the quickest way to give 50 people a chat window tomorrow.
2. Microsoft 365 Copilot
If your firm runs on Microsoft 365, Copilot is usually the easiest private option to justify. It works inside your tenant, follows your SharePoint and Teams permissions, and is covered by your existing Microsoft data protection terms (Microsoft). Permissions need attention first, because Copilot can surface any file a user already has access to. Review who can see restricted matters before rolling it out.
- Best for
- Firms already on Microsoft 365 that want one AI account across the firm
- Trains on your data?
- No. Prompts, responses and your Microsoft 365 data aren't used to train foundation models.
- Where data lives
- With your Microsoft 365 content, under its data residency commitments.
- Retention
- Admins set it in Microsoft Purview.
- Pricing
- Copilot Business £16.10 a user a month paid yearly, before VAT, on top of Microsoft 365. Microsoft shows a first-year promotional £13.80 until 31 December 2026 (Microsoft UK).
- Watch out for
- UK prompts can still be processed outside the UK. Microsoft now expects UK in-country processing by the end of 2026 (Microsoft). Anthropic's Claude models in Copilot fall outside those commitments and are off by default for UK tenants (Microsoft), so check before switching them on.
3. ChatGPT Enterprise and Business
The same ChatGPT your team already uses, on a business contract. Enterprise is the only ChatGPT plan that can store data at rest in the UK, an option OpenAI launched with the Ministry of Justice as its first customer (OpenAI). Business (formerly Team) has the same no-training terms without the residency option. For what ChatGPT is and isn't safe for in practice, see ChatGPT for lawyers.
- Best for
- Firms whose staff already use ChatGPT and need to bring it under a firm account
- Trains on your data?
- No, unless your firm opts in (OpenAI).
- Where data lives
- Enterprise: UK storage at rest for new workspaces. UK isn't offered for inference residency, which covers Europe, the US and the UAE (OpenAI). Business: no residency option.
- Retention
- Enterprise admins set it. Business keeps chats until deleted, then removes them within 30 days (OpenAI).
- Pricing
- Business is $20 a user a month billed annually, or $25 monthly (OpenAI). Enterprise is quote only.
- Watch out for
- It's still a standalone chat window, meaning staff copy client documents in and answers out, and governance relies on your AI policy.
4. Claude Team and Enterprise
Anthropic's business plans. Claude is strong with long documents and drafting, and the Commercial Terms bar training on your content (Anthropic). The main drawback for a UK firm is data location, as Anthropic's own platform only stores data in the US (Anthropic), meaning your DPIA has to cover a US transfer.
- Best for
- Firms that prefer Claude for drafting and are comfortable with a US transfer under the DPA
- Trains on your data?
- No. Covered by Anthropic's Commercial Terms.
- Where data lives
- US only, on Anthropic's own platform.
- Retention
- Chats kept until deleted, then removed within 30 days (Anthropic).
- Pricing
- Team from $20 a seat a month billed annually ($25 monthly), 2 to 150 seats. Enterprise $20 a seat a month billed annually with a 20-seat minimum, plus usage at API rates (Anthropic).
- Watch out for
- No UK or EU storage option on the Claude apps. For Claude in the UK, see the cloud tenant option below.
5. Harvey
A legal AI platform for research, review and drafting, built for large firms. Its security terms cover no training on customer data by default, zero data retention with model providers, and annual SOC 2 Type II and ISO 27001 audits (Harvey). I cover price and fit in Harvey AI pricing.
- Best for
- Large firms doing heavy review and drafting that want legal-specific workflows
- Trains on your data?
- No by default. A firm can opt in to a model trained only for it.
- Where data lives
- In-region in the EU or Switzerland, the US or Australia. No UK-only region listed.
- Retention
- Customers set retention and can delete data at any time.
- Pricing
- Not published.
- Watch out for
- No UK region, and the contracts are sized for large firms.
6. Legora
Harvey's closest rival, with a Word add-in and tabular review, and it sells to mid-sized firms too. Legora says customers control where their data is stored and it won't train on it (Legora security). Its hosting is "based primarily in Sweden, within the EU" (Legora). The two are compared in Legora vs Harvey.
- Best for
- Mid-sized and large firms that want a legal AI platform with EU hosting
- Trains on your data?
- No. Legora won't use your data to train or fine-tune models.
- Where data lives
- Mainly Sweden. EU-only can be configured. No UK region named publicly.
- Retention
- Set in your contract. Ask for it on the order form.
- Pricing
- Not published. You get a quote after a demo.
- Watch out for
- If a client needs UK-only hosting, get it in writing before you sign.
7. Lexis+ with Protégé
LexisNexis's AI research and drafting tool, grounded in its own legal sources. Of the legal tools here, it publishes the clearest retention rules: conversations are purged after 90 days or when the user deletes them, and uploaded documents after 10 minutes of inactivity (LexisNexis UK).
- Best for
- Research-heavy firms already paying for Lexis
- Trains on your data?
- No. LexisNexis and its AI providers are contractually barred from training on customer data.
- Where data lives
- Processed in the local region, listed as "Europe including UK" (LexisNexis UK). Ask which one applies to you.
- Retention
- Conversations up to 90 days. Uploaded documents 10 minutes after you stop using them.
- Pricing
- Not published. Quote from LexisNexis UK.
- Watch out for
- It's a research tool. It won't touch intake, email or billing.
8. Your own cloud tenant (AWS Bedrock London, Azure)
Frontier models run inside a cloud account your firm owns, in a region you choose. Claude Opus 5.5 now runs in-region on AWS Bedrock in London (AWS), keeping prompt processing in the UK. Azure has UK regions too, though newer OpenAI models are often only offered as Global or Data Zone deployments, which can process prompts outside the UK (Microsoft).
- Best for
- Firms with client contracts or policies requiring in-country UK processing as well as storage
- Trains on your data?
- No. The model maker doesn't see your prompts.
- Where data lives
- In the cloud region you choose, including London.
- Retention
- You set it, because it's your account.
- Pricing
- Pay per use at cloud rates, plus the cost of someone building and running it.
- Watch out for
- There's no chat app out of the box. Someone has to build what your team uses.
9. On-premise AI (Mistral, gpt-oss and other open models)
Running open-weight models on your own servers or private cloud gives the tightest control over data. Mistral says its enterprise assistant can run "self-hosted, in your public or private cloud" (Mistral), and OpenAI's gpt-oss models are free to download and run under Apache 2.0 (OpenAI).
- Best for
- Firms whose clients forbid any third-party processing
- Trains on your data?
- No. The data never reaches the model maker.
- Where data lives
- Your building or your private cloud.
- Retention
- Entirely yours.
- Pricing
- Mistral's enterprise plan is quote only. Self-hosting means buying or renting GPUs.
- Watch out for
- You run and patch the hardware, and open models usually trail the best hosted ones. For most firms, the infrastructure overhead outweighs the extra privacy benefit.
How to choose
Start with what fits your existing stack. For firms on Microsoft 365, Copilot is usually the most affordable compliant account across staff, while ChatGPT Enterprise makes sense if your team already prefers ChatGPT and needs UK storage.
Next, check your biggest clients' panel terms and contracts. If they require in-country UK processing, most hosted chat apps can't guarantee that today, pointing you towards your own cloud tenant in London or an on-premise setup.
Finally, look at where fee-earner hours are spent. A private chat window is safe, though staff still copy matter documents in and answers out. If most lost hours are spent on intake, email and client updates, that is what our AI operating system for law firms automates inside your existing systems, and AI for law firms explains how we set it up. Whatever you choose, document the approved tools and data rules in an AI policy.
