Keeping people safe starts with their information
The records an organisation keeps in CGRC AI are about people. Incident reports, safeguarding concerns, complaints and whistleblowing reports often hold the most sensitive things a patient, service user or member of staff will ever share. Protecting that information is part of keeping those people safe.
Sign-in and access
People sign in through a dedicated identity provider, which supports multi-factor sign-in and single sign-on with SAML or OpenID Connect.
What each person can see and do depends on their role and the Security Groups your organisation sets up. Every create, edit and delete is checked on the server, down to the individual record, and automated tests check these rules.
Each organisation's data is kept separate by design
Every business table carries an organisation identifier, and our coding rules require every query to be limited by it. Automated tests check this.
Audit trail and logging
Every create, update and delete is recorded in an audit log. Each organisation's log is hash-chained, which makes it tamper-evident. Our server logs carry identifiers and counts, never the content of a record.
AI that supports people
The AI drafts records and analysis. A person always reviews the draft and decides.
It works only from your organisation's own records and policies, showing which ones it used. It never sees another organisation's data.
Your organisation can switch the AI off altogether, or switch it off for a particular module or group of users.
How we build and check the software
Every change to the code runs automated tests. Every pull request that changes the main application's code is also checked by automated security scanning and an AI-assisted code review before it is merged. A person reviews the findings and decides. Our software libraries are checked weekly for known security advisories.
The application is built to send security headers that force secure connections and set a content security policy. The application's code uses no unsafe database queries or unsafe HTML insertion, and keys and other secrets are kept out of the code.
Backups and recovery
We have tested restoring a backup into a separate database, and every checksum matched. A written runbook sets out how we restore data. Pilot hosting will include automatic backups.
Hosting and data protection
Pilot environments will be hosted in a UK region on Microsoft Azure. The application is designed to force encrypted connections, and stored data will be encrypted at rest using the platform's standard encryption.
Cognito Consultants Ltd, which provides CGRC AI, is registered with the Information Commissioner's Office.
Some of our suppliers, including our sign-in and email providers, process data outside the UK under recognised safeguards. On request, we will share our list of sub-processors, where they process data and the safeguards in place, and complete your security questionnaire.
Reporting a security concern
If you think you have found a security problem, please email info@cgrcai.com and tell us how to see it. Please do not access or keep anyone else's data.
We have a written internal procedure for working through an incident and keeping a record. Breach notification times will be set out in the pilot agreement.
Questions organisations ask us
Does the AI keep a copy of our records?
Our AI service does not store your records. It logs only counts and error classes.
Can we limit a pilot to one service or team?
Yes. You can limit access through who is invited, job-title permissions and Security Groups.
Last updated: 20 October 2026
CGRC AI, from Cognito Consultants