Right First Time UK
Search Get Help
Home›UK Public Services & Transparency›UK Government: AI, Algorithms & Automation

UK Government: AI, Algorithms & Automation

AI and algorithms can help public bodies analyse information, improve services and reduce repetitive work. However, systems can also produce unfair, inaccurate or poorly explained outcomes when data, design, procurement, testing or human oversight are inadequate.

Use This Hub to Check the Evidence

This rebuilt page combines the newer Right First Time UK transparency layout with the detailed material from the original site. The original information, evidence routes and source links are kept below rather than being replaced.

1. Read the information

Use the restored sections below for the detailed subject guide and background.

2. Open the evidence

Source, evidence and information links are shown as clear buttons so you can check the underlying material.

3. Check current status

Where a subject changes over time, use the linked official source for the latest published position.

Information & Evidence Shortcuts

Quick access to source links already included in the original RFTUK page.

Open the Responsible AI ToolkitRead the Government AI PlaybookRead the Digital Government BlueprintVisit the Data Standards AuthorityOpen the Data and AI Ethics FrameworkRead the AI Assurance GuidanceSearch Official Transparency RecordsRead the Scope and Exemptions Policy

Full RFTUK Information & Evidence

◉ Find the system Identify which AI, algorithm, model or automated tool is being used.
⌘ Understand the decision Ask what data, rules and human checks contribute to the outcome.
✓ Check accountability Identify the responsible department, owner, supplier and reviewer.
£ Follow the money Search contracts, procurement records and departmental accounts.

What do AI and automated decision-making mean?

Different public bodies may use different terms. A system does not need to be described as “artificial intelligence” to have a significant effect on a person.

Artificial intelligence

Computer technology designed to perform tasks such as recognising patterns, producing content, classifying information, making predictions or supporting decisions.

Algorithm

A set of rules or calculations used by a computer system to process information and produce an output.

Automated decision

A decision made mainly or entirely by automated processing, potentially without direct human involvement.

Decision-support tool

Technology that recommends, scores or prioritises an outcome while a human officer is expected to make the final decision.

Risk scoring

A process that uses selected information to estimate the likelihood of fraud, harm, non-compliance, vulnerability or another event.

Machine learning

A method where a computer model learns patterns from data rather than relying only on manually written rules.

Generative AI

Technology that can generate text, images, audio, computer code or other content in response to instructions.

Profiling

Automated processing used to evaluate or predict aspects of a person’s behaviour, circumstances, performance, reliability or preferences.

The minimum accountability trail

Every significant public-sector AI or algorithmic system should have a clear trail from its purpose through to review and public challenge.

1 Purpose What problem is the system intended to solve?
2 Data What information is collected, matched or inferred?
3 Decision How does the output affect staff and the public?
4 Human control Can a trained officer examine and override the output?
5 Review Are errors, bias, complaints and outcomes monitored?

Official UK Government AI and data framework

These official resources set out expectations and guidance for responsible AI, digital government, data standards, assurance and algorithmic transparency.

✓

Responsible AI Toolkit

A government collection bringing together guidance intended to support responsible development and deployment of artificial intelligence.

  • AI assurance
  • Algorithmic transparency
  • Responsible innovation
  • Public attitudes and trust
AI

AI Playbook for the UK Government

Official guidance on the safe, effective and secure use of AI by civil servants and public-sector organisations.

  • Understanding capabilities and limitations
  • Selecting and purchasing AI
  • Safe deployment
  • Governance and human responsibility
◆

Blueprint for Modern Digital Government

The government’s blueprint covering digital public services, technology, data, public-sector reform and modernisation.

  • Digital public services
  • Cross-government technology
  • Data and AI
  • Skills and leadership
DS

Data Standards Authority

The official government body supporting consistent data standards and improved data use across public services.

  • Government data standards
  • Interoperability
  • Consistent data definitions
  • Cross-government adoption
DE

Data and AI Ethics Framework

Government guidance intended to help public-sector teams use data and AI responsibly, fairly and transparently.

  • Public benefit
  • Fairness
  • Transparency
  • Accountability
A

Introduction to AI Assurance

Official introductory guidance explaining how assurance methods can support the development and use of responsible AI.

  • Impact assessments
  • Audits and evaluations
  • Testing and monitoring
  • Independent assurance

Algorithmic Transparency Recording Standard

The Algorithmic Transparency Recording Standard provides a structured way for public bodies to publish information about algorithmic tools used in government and public services.

S

Mandatory Scope and Exemptions Policy

Explains the intended mandatory scope of the transparency standard and circumstances in which exemptions may apply.

G

Guidance for Organisations

Official instructions for organisations preparing and publishing an Algorithmic Transparency Recording Standard record.

Information to check in each algorithmic transparency record
  • Name of the public authority and responsible team
  • Name and description of the algorithmic tool
  • Policy purpose and intended public benefit
  • Whether the system makes or supports decisions
  • People, services or locations affected
  • Data collected, matched, generated or inferred
  • Legal basis for processing personal information
  • Supplier, developer and contractual arrangements
  • Testing for accuracy, bias and discriminatory effects
  • Human oversight and override arrangements
  • Appeal, correction and complaint routes
  • Monitoring, review and retirement date

Department-by-department transparency checks

These cards do not claim that every listed activity uses AI. They show the public-service areas where algorithms, data matching, automated processing or decision-support systems may require scrutiny.

AI procurement, suppliers and public contracts

Procurement transparency is essential because public bodies may buy software, cloud services, consultancy, licences, data analysis and automated-decision systems from outside suppliers.

£

Contracts Finder

Search public-sector contract opportunities, award notices and procurement information.

FT

Find a Tender

Search notices for higher-value public-sector procurement and contract awards.

MP

Government Major Projects Portfolio

Official information about major government projects, including delivery confidence and project data where published.

Contract information the public should check
  • Contracting authority and responsible senior officer
  • Supplier and any subcontractors
  • Original contract value and final expenditure
  • Software, cloud, consultancy and licence charges
  • Implementation and maintenance costs
  • Contract extensions and changes
  • Performance standards and service credits
  • Ownership of data, models and intellectual property
  • Exit arrangements and data deletion
  • Audit and inspection rights
  • Security requirements
  • Liability for inaccurate or discriminatory outcomes

Risks, safeguards and independent scrutiny

Responsible use requires more than purchasing a system. Public bodies must consider accuracy, data protection, equality, security, human responsibility and continuing monitoring.

!

Inaccurate outcomes

Incorrect, incomplete or outdated data may produce wrong recommendations, risk scores or decisions.

  • Test accuracy before deployment
  • Record known limitations
  • Measure false positives and false negatives
  • Provide correction and appeal routes
=

Bias and discrimination

Historical data or unsuitable design choices may disadvantage particular groups or reproduce existing inequalities.

  • Equality impact assessment
  • Testing across demographic groups
  • Monitoring real-world outcomes
  • Independent review where appropriate
D

Data protection

Personal data must be used lawfully, fairly, transparently and only for justified purposes.

  • Identify the lawful basis
  • Use data minimisation
  • Complete impact assessments
  • Protect access and retention
S

Cyber security

AI systems can introduce security risks involving models, data, suppliers, system access and malicious inputs.

  • Secure design and development
  • Supplier assurance
  • Logging and incident response
  • Secure maintenance and retirement
H

Automation bias

Staff may accept a computer recommendation without applying sufficient independent judgment.

  • Train staff on limitations
  • Provide meaningful override powers
  • Avoid performance targets that discourage review
  • Audit whether officers challenge outputs
R

Continuing review

A system that performed adequately during testing may behave differently as data, policy and public behaviour change.

  • Monitor outcomes and complaints
  • Retest after significant changes
  • Record incidents and corrections
  • Set a review or retirement date

Public rights and routes to challenge

The appropriate route depends on the organisation, the type of decision, the law governing the service and whether personal information was processed.

i

Ask for an explanation

Ask the authority whether an algorithm, risk model, automated process or AI tool contributed to the decision.

  • Name and purpose of the system
  • Information considered
  • Role of the automated output
  • Human checks applied
H

Request human review

Where relevant, ask for a suitably trained officer to review the information and reach a decision independently.

  • Correct inaccurate information
  • Provide missing evidence
  • Challenge unsuitable assumptions
  • Request written reasons
SAR

Subject access request

Ask an organisation for copies of personal information it holds about you, subject to the applicable legal rules and exemptions.

FOI

Freedom of Information request

Ask a public authority for recorded information about policies, contracts, assessments, audits or system performance, subject to lawful exemptions.

!

Make a complaint

Complain first to the organisation where appropriate. Keep copies of decisions, evidence, responses and relevant dates.

  • Explain the disputed outcome
  • Identify inaccurate information
  • Request correction and human review
  • Ask for the organisation’s final response
ICO

Complain to the ICO

The Information Commissioner may consider complaints about data protection or access-to-information matters.

Questions every public authority should answer

These questions can be used when examining a transparency record, privacy notice, contract, policy, impact assessment or Freedom of Information response.

What is the system called? Give the official name, version, supplier and internal project name.
What does it do? Explain the purpose and the output in clear language.
Who is affected? Identify the services, people, groups and geographic areas covered.
What data does it use? List the sources, categories, matching and inferred information.
Is personal information involved? State the lawful basis, retention period and data-sharing partners.
Does it make a decision? Explain whether it decides, recommends, scores, flags or prioritises.
Can a human disagree? Explain who can override the output and what evidence they receive.
How accurate is it? Publish testing methods, error rates and known limitations.
Was bias tested? Publish equality analysis and outcomes for affected groups.
What did it cost? Publish development, licence, cloud, maintenance and consultancy costs.
Who supplied it? Name all contractors, subcontractors and technology providers.
Who owns the data and model? Explain ownership, access, reuse and deletion arrangements.
What assessments were completed? Publish privacy, equality, security, ethics and algorithmic assessments.
What problems have occurred? Record incidents, complaints, corrections and service disruption.
What is the appeal route? Provide accessible instructions for review, correction and complaint.
When will it be reviewed? State the monitoring timetable, responsible owner and retirement plan.

Suggested Freedom of Information questions

Requests should be tailored to the authority and system concerned. Avoid asking for explanations or opinions where recorded information can be requested instead.

AI and algorithm system information

Please provide a list of artificial-intelligence, machine-learning, algorithmic, automated decision-making, risk-scoring and decision-support systems used by the authority.

For each system, please provide the name, purpose, operational status, responsible team, supplier, contract reference and the date it was first used.

Impact assessments and testing

Please provide copies of recorded impact assessments, equality assessments, data-protection impact assessments, security assessments, ethical assessments, validation reports and bias testing relating to the system.

Accuracy, errors and human overrides

Please provide recorded information showing the system’s measured accuracy, false-positive rate, false-negative rate, number of outputs reviewed by staff, number overridden by staff and number corrected following complaints or appeals.

Contracts and expenditure

Please provide the contract, award notice, original contract value, expenditure to date, extensions, change notices, implementation costs, annual licence costs, cloud costs, consultancy costs and maintenance charges relating to the system.

Algorithmic transparency record

Please confirm whether the authority has completed or published an Algorithmic Transparency Recording Standard record for the system. If one exists, please provide a copy or its published location.

Where no record has been completed, please provide recorded information explaining whether the authority considers the system outside the standard’s scope or subject to an exemption.

Official AI, algorithm and digital-government sources

Every link below goes to an official UK Government, regulator, parliamentary, audit or public-sector website.

GOV.UK — Responsible AI Toolkit GOV.UK — Blueprint for Modern Digital Government GOV.UK — Data Standards Authority GOV.UK — Algorithmic Transparency Records GOV.UK — ATRS Mandatory Scope and Exemptions GOV.UK — Guidance for Organisations Using ATRS GOV.UK — Algorithmic Transparency Template GOV.UK — AI Playbook for the UK Government GOV.UK — AI Resources for the Public Sector GOV.UK — AI Insights GOV.UK — Introduction to AI Assurance GOV.UK — Assuring a Responsible Future for AI GOV.UK — Data and AI Ethics Framework GOV.UK — Government Service Manual GOV.UK — Government Service Standard GOV.UK — Technology Code of Practice GOV.UK — Government Major Projects Data Contracts Finder — Search Public Contracts Find a Tender — Search Procurement Notices Digital Marketplace ICO — Artificial Intelligence Guidance ICO — AI and Data Protection ICO — AI and Data Protection Risk Toolkit ICO — Explaining Decisions Made with AI NCSC — Secure AI System Development NCSC — Artificial Intelligence Security Guidance UK Government Security — Artificial Intelligence National Audit Office — AI Reports National Audit Office — Digital Transformation Reports UK Parliament — Committees UK Parliament — Search AI in Hansard ICO — Make a Complaint

Disclaimer: This page provides general public-interest information and signposting. It is not legal, data-protection, technical or cyber-security advice. The use of AI, algorithms or automation does not by itself prove wrongdoing, unfairness or unlawful decision-making. Systems, policies and official guidance may change. Readers should examine the latest official publication, relevant legislation and the facts of each individual case.

← All Transparency HubsInformation & Guides