The shared challenge
Regulated AI asks the same three questions.
Financial services and healthcare differ in their rules, not in what a reviewer wants to see.
01 / Accountability
Financial servicesNamed senior individuals are accountable for AI-assisted decisions.
HealthcareDCB0129 (for manufacturers) and DCB0160 (for deploying organisations) set clinical risk management expectations for health IT.
02 / Evidence trails
Financial servicesA regulator can ask how one decision was made.
HealthcareClinical safety documentation, data protection impact assessments (DPIAs) and, for suppliers, DTAC all depend on clear records of what a system does and why.
03 / Human oversight
Both sectorsA person should be able to review, challenge and override an AI-assisted output, and you should be able to show that they did.
For public-sector transparency, the Algorithmic Transparency Recording Standard (ATRS) is the UK government’s standard for publishing information about algorithmic tools. It is mandatory for central government departments and certain arm’s-length bodies, and recommended more widely across the public sector.
Sources: DCB0129 and DCB0160 (NHS England); DTAC (NHS AI and Digital Regulations Service); DPIAs (Information Commissioner’s Office); ATRS (GOV.UK). This is a comparison of approaches, not legal advice. Standards change, so check the current version of each.
How we work
Controls built into systems,
not only written in policy.
A policy says what should happen. A control makes it happen, and leaves a record. It is the approach behind ComplyAI, our financial services system.
01 / Verified evidence
Every claim is traced to a source you can inspect, and every quotation is checked against that source.
02 / Human review
A person reviews each record before it is relied on.
03 / Fail-closed checks
If a check cannot confirm something, the answer is no, not “probably”.
04 / Audit trails
A record of who decided what, when and why.
Research
Evidence, gathered carefully.
Health AI Governance Evidence Index (working title)
We are building a way to gather public evidence of how health organisations govern AI, and to keep that evidence accurate. We have not published any findings.
- Public sources only. robots.txt is respected.
- Every quotation is checked against its source.
- A person reviews every record before it is used.
- Organisations, not people. The index does not record individuals.
- Corrections are welcome.
An independent project, not affiliated with or endorsed by the NHS.
How the crawler worksRegister your interest in the index
We are planning the following. None of it is available yet, and the details may change.
- Free tracker. Planned, not yet available.
- Pro, from £149 to £249 a year. Planned, not yet available. Indicative price only.
- Annual report. Planned, not yet available.
- Trust dossiers. Planned, not yet available.
To be told when the index opens, email hello@shiftaiconsulting.co.uk with your name, organisation and the options you are interested in.
By emailing us you consent to ShiftAi Systems Ltd using your name, organisation and email address to tell you about the Evidence Index. You can withdraw consent at any time by emailing us. See our privacy notice.
How we can help
Support for the people
who hold the accountability.
Trusts and ICBs
Governance you can show to a reviewer
- An AI governance review.
- AI policy and oversight design.
- Support to help you prepare evidence for clinical safety and DPIA processes.
Suppliers
Governance that stands up to scrutiny
- AI governance support for your product and your team.
- Help preparing for DTAC (the Digital Technology Assessment Criteria).
We support your teams. We do not certify or approve anything. Clinical safety cases, DPIAs and DTAC submissions stay with your accountable people.
About
Built by an engineer. Held to a research standard.
Ifat Noreen founded ShiftAi Systems Ltd after nine years of computational science research at the University of Southampton. In that world, a result nobody can reproduce does not count.
We hold AI governance to the same standard: if a decision cannot be reconstructed and evidenced, it is not governed. ShiftAi is an independent company and is not affiliated with or endorsed by the NHS.