Bank-grade security and valid insurance, built in from day one.
Whatever your business, two immediate risks arise when adopting AI: confidential records leaking through shared cloud servers or messy internal drives, and unsupervised AI outputs invalidating your Professional Indemnity or commercial insurance cover.
Sentivox removes both risks by design. Your workspace and AI run inside self-sterilising UK clean rooms locked under your own master key, while built-in human checkpoints and a 7-year flight recorder keep every piece of work verifiable, supervised, and fully insurable.
Four Layers of Protection.
How we protect your confidential files at every stage—without requiring your business to manage servers or hire security engineers.
Scrambled Before It Leaves Your Screen
When your team asks a question or uploads a document, it is scrambled directly inside their web browser before travelling across the network. Plain text only ever exists on your staff’s screens and inside your private UK clean room.
The Self-Sterilising UK Clean Room
Your AI runs inside a sealed hardware clean room in London with no outbound internet connection. While a task runs, the physical processor chip locks working memory so nobody—not even data-centre technicians or Sentivox engineers—can look inside. The microsecond the task finishes, that memory is wiped clean.
Your Business Holds the Master Key
Your saved vaults, finished work, and audit logs sit in a dedicated UK database locked under a master key held inside your own Microsoft or Google account. Neither Sentivox nor any cloud provider holds a copy—and if you ever turn your key off, your stored files become unreadable immediately.
The Executive Sign-Off Rule
Sentivox agents never act behind your back or issue unsupervised external advice. Whenever an agent checks accounts, reviews a contract pack, or prepares an action, it presents a clear summary linked to the exact source page and waits for a qualified team member to click approve.
Why Public Cloud AI Leaves You Exposed—And How Sentivox Locks the Perimeter.
Even on corporate plans, standard AI tools create internal permission leaks, allow accidental uploads to outside models, and harvest staff usage telemetry. We built three structural locks to prevent this.
Role-Restricted Vaults vs. Messy Shared Drives
Public tools like Microsoft Copilot index your entire SharePoint or shared drive at once. If a sensitive HR, partner-pay, or M&A folder was accidentally left open years ago, any junior employee can surface it with a single prompt. Sentivox eliminates this blast radius completely: you place files into deliberate, role-restricted vaults synced with your Microsoft 365 or Google security groups.
Automatic Vault Lock When Using Public AI Models
Sometimes staff want to use a frontier public model (such as GPT-5, Claude, or Gemini) for general research or creative drafting. In Sentivox, the moment a user switches from your Sovereign UK Model to a public model, our Automatic Vault Lock physically disconnects your document vaults and disables file uploads—making it impossible to accidentally paste or attach confidential client records to an outside server.
Zero Telemetry & Zero Snooping on Your Staff
Big Tech cloud platforms operate like nosey landlords—quietly collecting usage telemetry, prompt metadata, and behavioural analytics on how your staff work. Sentivox operates like an old-fashioned 1990s commercial landlord: we provide the reinforced building, maintain the perimeter, and hand you the keys. We never read your documents, never monitor your prompts, and collect zero behavioural telemetry.
| Security & Governance Dimension | Public Cloud AI (Even on Corporate Terms) | Sentivox Private UK Clean Rooms |
|---|---|---|
| Internal File Blast Radius | Searches across entire legacy SharePoint drives—risking staff discovering partner pay, HR files, or restricted deals | Role-restricted vaults synced to your security groups; staff only ever access folders their role explicitly permits |
| Using Public AI Models Safely | Staff can freely upload confidential client PDFs and spreadsheets into public chat windows | Automatic Vault Lock physically disconnects client vaults and blocks uploads whenever a public model is selected |
| Where Files Are Processed | Processed on shared multi-tenant cloud servers connected to the open internet | Processed inside a self-sterilising London clean room with zero outbound internet connection |
| Who Holds the Master Key | The AI vendor holds the encryption keys and retains administrative control over your history | Your business holds the sole master key in your own Microsoft or Google account with an instant off-switch |
| Usage Telemetry & Snooping | Vendor collects metadata, usage patterns, and diagnostic telemetry across your workforce | 1990s Honest Landlord Covenant: zero behavioural telemetry, zero prompt inspection, zero data harvesting |
| Fact-Checking & Insurability | Unverified answers and autonomous actions that risk triggering 2026 Absolute AI Insurance Exclusions | Exact page-and-line citations, mandatory human sign-off checkpoints, and a 7-year insurer evidence pack |
Built for the 2026 Insurance & Regulatory Reality.
Protecting your data is only half the battle. In 2026, insurers and UK regulators require provable human supervision over how AI is tested, signed off, and audited.
Following January 2026 insurer policy shifts (including Verisk’s Absolute AI Exclusions), commercial and PII underwriters increasingly refuse cover for claims arising from unverified AI hallucinations or autonomous, unsupervised AI actions.
An October 2026 University of Cambridge study of 1,027 UK leaders found that while ~70% of firms have piloted AI, full rollout stalls on commercial liability, data security, and proving human supervision under audit.
Across the FCA, ICAEW, RICS, SRA, and corporate governance standards, 2026 UK guidance mandates that senior decision-makers remain accountable—requiring clear source traceability and recorded human sign-off.
How We Keep Every Workflow Supervised and Insured
Shadow Mode Dry-Run Testing
Before any background agent or workflow goes live across your team, it runs in Shadow Mode alongside your existing staff processes. You can compare the AI’s draft findings against your team’s actual completed work with zero operational exposure—signing off for live deployment only when accuracy is proven.
Proportionate Human Checkpoints
Under 2026 insurance underwriting rules, unverified AI output risks falling into an uninsured exclusion zone. Sentivox enforces proportionate human checkpoints: the AI performs the heavy reading, cross-checking, and drafting with clickable page-and-line citations, while a qualified human clicks to verify and sign off—legally converting the output into a supervised, insured professional act.
7-Year Flight Recorder & 1-Click Evidence Pack
Every document version, prompt, source citation, and human approval timestamp is written to a tamper-proof 7-year compliance log in London. At renewal time or during an audit, your Compliance Officer or Finance Director can export a One-Click Annual Insurance & Audit Evidence Pack to hand straight to your broker or regulator.
Six Questions Asked by Managing Partners, Brokers, and IT Leads
Does our confidential work or client data ever train Sentivox or outside AI models?
Never—both contractually and by physical design. Your Sovereign AI models run inside a self-sterilising UK clean room with zero outbound connection to the public internet, and working memory wipes itself the instant each task finishes. And if a staff member switches to a public model for general research, our Automatic Vault Lock physically disconnects your client vaults so confidential files can never leak out.
How does Sentivox protect our Professional Indemnity Insurance (PII) against 2026 Absolute AI Exclusions?
Since January 2026, commercial underwriters (following verdicts such as Verisk’s Absolute AI Exclusions) have moved to exclude liabilities caused by unsupervised or hallucinated AI outputs. Sentivox keeps your cover intact by ensuring AI never acts autonomously on material tasks: every finding links to the exact source page, mandatory human checkpoints convert AI preparation into a supervised human act, and our One-Click Annual Insurance & Audit Evidence Pack proves your governance to underwriters at renewal.
Can Sentivox engineers or data-centre technicians ever read our files or monitor how we work?
No. Under our 1990s Honest Landlord Covenant, we collect zero usage telemetry and cannot inspect your files. Your documents are scrambled in the browser, locked under a master key held in your own firm’s cloud account, and processed inside hardware-locked chip memory. Even Sentivox’s own engineers only ever see encrypted ciphertext—preserving legal privilege, fiduciary duty, and commercial secrecy.
How do you stop a junior employee from accidentally uncovering partner pay, HR files, or restricted deals?
Unlike Microsoft Copilot—which searches across your entire legacy SharePoint drive and exposes any folder with messy historical permissions—Sentivox operates with a Zero SharePoint Blast Radius. Documents are placed into clean, role-restricted vaults synced directly with your Microsoft 365 or Google security groups. If a team member is not cleared for a vault, neither they nor their AI assistant can ever see or query it.
How does Sentivox meet 2026 UK regulatory supervision mandates and UK GDPR rules?
An October 2026 University of Cambridge study of 1,027 UK decision-makers found that while ~70% of organisations have piloted AI, production deployment stalls on commercial liability, data security, and the inability to prove human supervision under audit. Sentivox closes this gap completely: 100% of your clean room, database, and 7-year flight recorder remain strictly in London (UK), Shadow Mode validates workflows before go-live, and every human sign-off is logged for regulatory inspection.
What happens if we ever need to lock the system immediately or leave Sentivox?
Your business holds the ultimate physical off-switch. Because your master encryption key lives inside your own Microsoft Azure or Google Cloud account, your IT team can revoke our access in a single click. The moment you disable your key, all stored vaults and archives become permanently unreadable to everyone—including Sentivox.
Walk through our security and insurance architecture with a senior architect.
Invite your Compliance Officer, Finance Director, insurance broker, or outsourced IT provider (MSP) to a 30-minute scoping briefing. We will walk you through our UK clean rooms, customer-held master keys, and One-Click Annual Insurance & Audit Evidence Pack in plain English.