Trust centre
We build AI tools for global brands and FTSE 100 businesses. This is how we keep your data, your systems and your reputation safe.
Request our security packYour cloud or ours, in the region you choose. Clients can host our tools themselves.
Our AI runs on Anthropic's business API: no training, prompts deleted within 30 days.
Under our current terms, if anything touches your data, you hear from us within a day.
No servers, no office network. Encrypted laptops, MFA on every staff account, automatic updates.
A person on our team reviews AI output in everything we deliver or publish. Tools we build include review steps you control.
Every new build starts with an exit plan: who runs it, what it costs, what to watch.
For your security, privacy and procurement teams.
Multi-factor authentication required for every user through Microsoft 365 security defaults. Administrator rights are held by one named person.
Every company laptop is encrypted (BitLocker on Windows, FileVault on Mac), installs updates automatically and runs built-in antivirus.
Microsoft 365, Azure and Google Cloud, encrypted in transit and at rest. Our providers hold ISO 27001 and SOC 2 Type II.
Private GitHub repositories, changes reviewed before release, automated tests, and dependency alerts on every repository.
Sign-ins, permission changes and sharing logged in Microsoft 365 and kept for 180 days.
A written incident procedure, 24-hour notice to clients under our current terms, and a continuity plan tested every year, most recently in August 2026.
Security and data protection training, recorded in a training register.
Penetration testing of a tool we build for you can be arranged on request, by your own security team or an independent tester.
Around twenty written security, privacy and AI policies, following the ISO 27001 structure, written in 2025 and 2026 and reviewed yearly. A risk register, last fully assessed in January 2026.
Critical security fixes within 14 days, high within 30. Laptops update automatically, and every code repository has dependency alerts switched on.
Suppliers that touch client data are checked for certifications, data protection terms and data location when added, and every year from 2027. Every one is listed in the Sub-processors tab.
Our London office is in a managed building with controlled entry and CCTV in shared areas. No client data is stored on site.
Claude, through Anthropic's business API. Anthropic holds ISO 27001, ISO 42001 and SOC 2 Type II.
Never on your data. Anthropic does not train on API data, and neither do we, unless you ask us to for your own tool.
Anthropic deletes prompts and outputs within 30 days. Zero data retention can be arranged for your project.
A person approves every output before it is delivered, published or sent on your behalf.
Tell us not to use AI on all or part of your work, at any time. We apply it within five working days.
What we make with AI is yours, on the same terms as all our work.
A tool sends the AI only the text a task needs. Client data stays in the systems and database the client chooses.
In our reference design, every request is checked against the user's own organisation before data is read. AI keys are kept encrypted, never in code, and never shared between clients.
Our full notice: fiftyfiveandfive.com/privacy.
Ready to sign with your work order: 24-hour breach notice, 30 days' notice of new sub-processors.
Access, correction or deletion, free, at privacy@fiftyfiveandfive.com. Answered within a month.
No cookies for visitors, no ad trackers. Our own analytics honours Global Privacy Control.
Contacts deleted three years after last contact. Website analytics kept 14 months.
Chris Wright, Founder, with a named deputy.
Where a supplier is outside the UK, a UK-approved safeguard applies to every transfer. Each one is named in the Sub-processors tab.
For our own marketing we hold business contact data only, never special category data, and we tell people how to object or opt out in every email.
| Supplier | What for | Where | Safeguard |
|---|---|---|---|
| Microsoft 365 | Email, files, Teams | United Kingdom (data at rest) | UK data residency; UK Extension to the Data Privacy Framework, plus UK Addendum to the SCCs |
| Microsoft Azure | Hosting for some client tools | Region chosen per project | UK Extension to the Data Privacy Framework, plus UK Addendum to the SCCs |
| GitHub | Code repositories | United States and other locations | UK Extension to the Data Privacy Framework, plus UK Addendum to the SCCs |
| Google Cloud | Hosting for some client tools | Region chosen per project | UK Extension to the Data Privacy Framework, plus UK Addendum to the SCCs |
| Google Firebase | Website hosting and our own analytics | Global network; analytics data in the United States | UK Extension to the Data Privacy Framework, plus UK Addendum to the SCCs |
| OVHcloud | Hosting for our CRM | France | UK adequacy (EEA) |
| Anthropic | AI (Claude API, business terms) | Stored in the United States; may be processed in other regions | UK Addendum to the SCCs |
| Notion | Internal documents | United States | UK Addendum to the SCCs |
| Adobe | Contract signing (Acrobat Sign) | European Union or United States | Contract with Adobe Ireland; UK Extension to the Data Privacy Framework |
| Mailchimp | Newsletter | United States and other locations | UK Extension to the Data Privacy Framework, plus UK Addendum to the SCCs |
| Apollo | Business contact data (independent controller) | United States | UK Extension to the Data Privacy Framework |
| Stripe | Online payments (independent controller) | United States and other locations | UK Extension to the Data Privacy Framework, plus UK Addendum to the SCCs |
| PayPal | Online payments (independent controller) | United Kingdom, United States and other countries | PayPal UK Ltd; contract terms for transfers |
We give clients 30 days' notice before adding or replacing a sub-processor. Last changed 17 November 2026.
Our platforms do: Microsoft, Google and Anthropic hold ISO 27001 and SOC 2 Type II. Our own policies follow the ISO 27001 structure, and we are working towards Cyber Essentials and ISO 27001 ourselves. A FTSE 100 client completed a full supplier security assessment of us in 2026.
Our Founder, with a second approver for policy changes. Every project has a named person accountable for its data.
Yes. Around twenty, covering acceptable use, access control, data protection, incidents, continuity, secure development, suppliers, AI and more. Written in 2025 and 2026 and reviewed yearly. Available on request.
A risk register, assessed and reviewed by management. The latest full assessment was in January 2026.
Staff complete security and data protection training, recorded in a training register.
Yes. Microsoft 365 security defaults require multi-factor authentication for every user, and only one person holds administrator rights.
Every company laptop is encrypted (BitLocker on Windows, FileVault on Mac), installs updates automatically and runs built-in antivirus: Microsoft Defender Antivirus on Windows, the built-in malware protection on Mac.
Yes, in transit and at rest, across Microsoft 365, Azure and Google Cloud.
Wherever you choose: in your own cloud, or in ours on Azure or Google Cloud, in the region you pick. Clients can host our tools themselves.
We follow a written incident procedure and, under our current terms, tell you within 24 hours if your data is affected. Microsoft and Google run 24-hour security operations on the platforms underneath.
Yes. We test it every year with all staff, most recently in August 2026. With no office servers, losing a building or a laptop does not stop the work.
Code lives in private GitHub repositories. Changes are reviewed before release, tests run automatically, and every repository has dependency alerts switched on.
Only the suppliers listed on this page. Those that process data for us work under written data protection terms; Apollo, Stripe and PayPal act as independent controllers under their own terms. We give 30 days' notice before adding one.
No. Our AI runs on Anthropic's business API, which does not train on customer data, and we never train models on your data unless you ask us to for your own tool.
Every tool runs automated tests before each release, AI features are checked against questions with known answers, and you sign off each stage before the next starts. A person on our team reviews AI output in everything we deliver.
You do, outright, unless we agree at the start that a tool is ours and licensed to you. It is in the contract either way.
Yes. Every new build starts with an exit plan, and we hand over code, documentation and a runbook so your team or another partner can run it.
Yes, once a year as standard. Ask for our completed one first: it usually answers everything.
Yes, under our data processing agreement.
How we build
The same production line for every tool, so you always know what happens next, who signs it off, and how you would run it without us.
Agree
Recorded working sessions turn into a written spec, an ownership schedule and the exit plan.
Gate: you sign the spec and exit plan before any code is written.
The architecture in your terms, a data-flow diagram and a security pack for your review, plus the screens.
Gate: your technology team signs the architecture.
Build
From our own proven patterns, never another client's work, in the repository the contract names.
Gate: every change reviewed and tested before it goes in.
Automated tests with the code, and AI features checked against questions with known answers.
Gate: tests pass before release.
Into your cloud or ours, versioned, with release notes every time.
Gate: a tested way to roll back.
A runbook written from the work itself: how to run, watch and fix it.
Gate: your team can run it from the runbook.
Hand over
Training on your own data, with us alongside your team for the first weeks.
Gate: your people use it unaided, with a named owner.
Monitoring, patching, access reviews, and a yearly rehearsal of the exit plan.
Gate: your engineer ships a change while we watch.
On-request documents arrive as view-only links, usually the same working day.
Tell security@fiftyfiveandfive.com. We reply within two working days and never act against good-faith research.