[] CONTENTS 1Purpose of this Act 2Systems to which this Act applies 3Algorithmic Impact Assessments 4Algorithmic Transparency Records 5Requirements of public sector organisations on use of algorithmic or automated decision-making systems 6Training of public sector employees 7Logging 8Prohibition on procuring systems incapable of scrutiny 9Independent dispute resolution service 10Regulations 11Definitions 12Extent, commencement and short title Schedule 1Principles for reviewing, explaining and overseeing operation of algorithmic and automated decision-making systems Schedule 2Definitions used in this Act []

A

bill

to

Regulate the use of automated and algorithmic tools in decision-making processes in the public sector; to require public authorities to complete an impact assessment of automated and algorithmic decision-making systems; to ensure the adoption of transparency standards for such systems; and for connected purposes.

Be it enacted by the King’s most Excellent Majesty, by and with the advice and consent of the Lords Spiritual and Temporal, and Commons, in this present Parliament assembled, and by the authority of the same, as follows:—
1 Purpose of this Act

The purpose of this Act is to ensure that algorithmic and automated decision-making systems are deployed in a manner that accounts for and mitigates risks to individuals, public authorities, groups and society as a whole, and leads to efficient, fair, accurate, consistent, and interpretable decisions; and to make provision for an independent dispute resolution service.
2 Systems to which this Act applies (1)

Subject to subsections (3) and (4) this Act applies to any algorithmic or automated decision-making system developed or procured by a public authority from six months after the date on which this Act is passed.

 (2)

This includes—

 (a)

any system, tool or statistical model used to inform, recommend or make an administrative decision about a service user or a group of service users, and

 (b)

systems in development, excluding automated decision-making systems operating in test environments.

 (3)

This Act does not apply to any automated decision-making system used for the purpose of national security.

 (4)

This Act does not apply to automated systems which merely calculate and implement formulas, including taxation and budgetary allocation, insofar as they automate a process of calculation which would otherwise be carried out manually and fully understood.
3 Algorithmic Impact Assessments (1)

Prior to deployment of an algorithmic or automated decision-making system, public authorities are responsible for completing an Algorithmic Impact Assessment prescribed in regulations made under this Act.

 (2)

Subsection (1) does not apply when the algorithmic or automated decision-making system is—

 (a)

used solely for the formulation of policy in relation to that public authority, and

 (b)

is not expected to, in practice, fully or predominantly determine the content of the policy.

 (3)

The Algorithmic Impact Assessment must be updated when the functionality, or the scope, of the algorithmic or automated decision-making system changes.

 (4)

The final Algorithmic Impact Assessment must be published in accessible format within 30 days of the results being known.

 (5)

The Secretary of State must by regulations prescribe the form of an Algorithmic Impact Assessment framework with the aims of ensuring public authorities—

 (a)

procure, develop, and implement algorithmic and automated decision-making systems such that the decisions made in and by a public authority are responsible and comply with procedural fairness and due process requirements, and its duties under the Equality Act and the Human Rights Act 1998,

 (b)

assess the impacts of algorithms on administrative decisions, minimise negative outcomes, and evaluate the potential to maximise positive outcomes,

 (c)

make data and information on the use of algorithmic and automated decision-making systems in public authorities available to the public,

 (d)

better understand and reduce the risks associated with algorithmic and automated decision-making systems,

 (e)

introduce the appropriate governance, oversight, and reporting and auditing requirements that best match the risks associated with the application envisaged, and

 (f)

undergo responsible innovation of algorithmic and automated decision-making systems.

 (6)

The framework as prescribed by regulations made under subsection (5) must include the requirement for—

 (a)

a detailed description of the algorithmic or automated decision-making system,

 (b)

an assessment of the relative benefits and risks of the system including the risks to the privacy and security of personal information, risks to the safety of a service user or group of service users, and risks and likely impacts on employees of public authorities;

 (c)

an explanation of the steps taken to minimise those risks,

 (d)

independent external scrutiny of the efficacy and accuracy of the system, and

 (e)

mandatory bias assessment of any algorithmic or automated decision-making system to ensure it abides by the Equality Act and the Human Rights Act 1998.

 (7)

The Secretary of State must publish regulations made under subsection (5) in draft and consult such persons they consider appropriate on the draft regulations before laying the regulations before both Houses of Parliament.
4 Algorithmic Transparency Records (1)

Prior to use or procurement of an algorithmic or automated decision-making system, public authorities must complete an Algorithmic Transparency Record prescribed in regulations made under this Act.

 (2)

Subsection (1) does not apply when the algorithmic or automated decision-making system is—

 (a)

used solely for the formulation of policy in relation to that public authority, and

 (b)

is not expected to, in practice, fully or predominantly determine the content of the policy.

 (3)

The Algorithmic Transparency Record must be published in accessible format within 30 days of the completion of the record.

 (4)

The Algorithmic Transparency Record must be updated when the functionality, or the scope, of the algorithmic or automated decision-making system changes.

 (5)

The Secretary of State must by regulations prescribe the form of transparency records with the aim of ensuring public authorities increase the transparency of algorithm-assisted decisions.

 (6)

The Algorithmic Transparency Record as prescribed by regulations made under subsection (1) must include the requirement for—

 (a)

a detailed description of the algorithmic or automated decision-making system,

 (b)

an explanation of the rationale for using the system,

 (c)

information on the technical specifications of the system,

 (d)

an explanation of how the system is used to inform administrative decisions concerning a service user or group of service users, and

 (e)

information on human oversight of the system.
5 Requirements of public sector organisations on use of algorithmic or automated decision-making systems (1)

No later than the commencement of use of a relevant algorithmic or automated decision-making system, a public authority must—

 (a)

give notice on a public register that the decision rendered will be undertaken in whole, or in part, by an algorithmic or automated decision-making system,

 (b)

make arrangements for the provision of a meaningful and personalised explanation to affected individuals of how and why a decision affecting them was made, including meaningful information about the decision-making processes, and an assessment of the potential consequences of such processing for the data subject, as prescribed in regulations to be made by the Secretary of State,

 (c)

develop processes to—

 (i)

monitor the outcomes of the algorithmic or automated decision-making system to safeguard against unintentional outcomes and to verify compliance with this Act and other relevant legislation, and

 (ii)

validate that the data collected for, and used by, the system is relevant, accurate, up-to-date, and in accordance with the Data Protection Act 2018, and

 (d)

make arrangements to conduct regular audits and evaluations of algorithmic and automated decision-making systems, including the potential risks of those systems and steps to mitigate such risks, as prescribed in regulations to be made by the Secretary of State.
6 Training of public sector employees (1)

Public authorities using an algorithmic or automated decision-making system to inform or recommend an administrative decision concerning a service user or group of service users must implement organisational practices and measures to ensure that those applying the final decision have the authority and competence to challenge the system’s output.

 (2)

Each public authority that uses an algorithmic or automated decision-making system must provide adequate employee training in the design, function, and risks of the system, in order to review, explain and oversee its operations in accordance with the principles set out in Schedule 1.
7 Logging (1)

All algorithmic and automated decision-making systems must be designed with logging capabilities enabling the automated recording of events during operation, in line with recognised standards or common specifications, which enable the monitoring of the operation of the system in relation to risks and legal obligations.

 (2)

Logs referred to in subsection (1) must be held by, or regularly transmitted to, the public authority with responsibility for the functions being exercised in connection with the algorithmic or automated decision-making system.

 (3)

Public authorities must hold logs for a minimum period of five years, unless a shorter period is strictly necessary for purposes of privacy or security, such period to be determined in advance of the use of the algorithmic or automated decision-making system.

 (4)

In the case of decision support systems, logs referred to in subsection (1) must record whether or not the final decision taken followed the recommendation of the algorithmic or automated decision-making system.
8 Prohibition on procuring systems incapable of scrutiny (1)

No public authority shall deploy or use an algorithmic or automated decision-making system where there are practical barriers, including contractual or technical measures and intellectual property interests, limiting their effective assessment or monitoring of the algorithmic or automated decision-making system in relation to individual outputs or aggregate performance.

 (2)

In assessing their obligations under subsection (1), public authorities should require vendors of algorithmic or automated decision-making systems to—

 (a)

disclose the results of evaluations carried out on those systems, including evaluations of foundation models used as components within the system, and

 (b)

on request, submit systems and relevant documentation to the AI Safety Institute for evaluation.
9 Independent dispute resolution service

The Secretary of State must ensure that the ability to—

 (a)

challenge a decision or class of decisions made by an algorithmic or automated decision-making system, or

 (b)

obtain redress for a decision or class of decisions made by an algorithmic or automated decision-making system

is available through an independent dispute resolution service appropriate to the nature of the public authority and the decision or class of decisions in question.
10 Regulations

Regulations under sections 3, 4 and 5 are to be made by statutory instrument and may not be made unless a draft has been laid before, and approved by a resolution of, both Houses of Parliament.
11 Definitions

Schedule 2 contains definitions of terms used in this Act.
12 Extent, commencement and short title (1)

This Act extends to England and Wales.

 (2)

This Act comes into force six months after the day on which it is passed.

 (3)

This Act may be cited as the Public Authority Algorithmic and Automated Decision-Making Systems Act 2024.
Schedules Schedule 1

Section 6

 Principles for reviewing, explaining and overseeing operation of algorithmic and automated decision-making systems 1

Employees in public authorities which use algorithmic or automated decision-making systems must review, explain and oversee operations in accordance with the principles and their interpretation as set out in paragraphs 3 to 14.

 2

The principles are complementary and should be considered as a whole.

 Inclusive growth, sustainable development and well-being 3

Stakeholders should proactively engage in responsible stewardship of trustworthy AI in pursuit of beneficial outcomes for people and the planet, such as—

 (a)

augmenting human capabilities and enhancing creativity,

 (b)

advancing inclusion of underrepresented populations,

 (c)

reducing economic, social, gender and other inequalities, and

 (d)

protecting natural environments.

in order to support inclusive growth, well-being, sustainable development and environmental sustainability.

 Respect for the rule of law, human rights and democratic values, including fairness and privacy 4

AI actors should respect the rule of law, human rights, democratic and human-centred values throughout the AI system lifecycle, including—

 (a)

non-discrimination and equality,

 (b)

freedom,

 (c)

dignity,

 (d)

autonomy of individuals,

 (e)

privacy and data protection,

 (f)

diversity,

(g)

fairness,

 (h)

social justice, and

 (i)

internationally recognised labour rights.

 5

Misinformation and disinformation amplified by AI should be addressed while respecting freedom of expression and other rights and freedoms protected by applicable international law.

 6

AI actors should implement mechanisms and safeguards, such as capacity for human agency and oversight, including to address risks arising from uses outside of intended purpose, intentional misuse, or unintentional misuse in a manner appropriate to the context and consistent with the state of the art.

 Transparency and explainability 7

AI Actors should commit to transparency and responsible disclosure regarding AI systems, meaning they should provide meaningful information, appropriate to the context, and consistent with the state of art—

 (a)

to foster a general understanding of AI systems, including their capabilities and limitations,

 (b)

to make stakeholders aware of their interactions with AI systems, including in the workplace,

 (c)

where feasible and useful, to provide plain and easy-to-understand information on the sources of data/input, factors, processes and/or logic that led to the prediction, content, recommendation or decision, to enable those affected by an AI system to understand the output, and,

 (d)

to provide information that enable those adversely affected by an AI system to challenge its output.

 Robustness, security and safety 8

AI systems should be robust, secure and safe throughout their entire lifecycle so that, in conditions of normal use, foreseeable use or misuse, or other adverse conditions, they function appropriately and do not pose unreasonable safety and/or security risks.

 9

Mechanisms should be in place, as appropriate, to ensure that if AI systems risk causing undue harm or exhibit undesired behaviour, they can be overridden, repaired, and/or decommissioned safely as needed.

 10

Mechanisms should, where technically feasible, be in place to bolster information integrity while ensuring respect for freedom of expression.

 Accountability 11

AI actors should be accountable for the proper functioning of AI systems and for the respect of the above principles, based on their roles, the context, and consistent with the state of the art.

 12

To this end, AI actors should ensure traceability, including in relation to datasets, processes and decisions made during the AI system lifecycle, to enable analysis of the AI system’s outputs and responses to inquiry, appropriate to the context and consistent with the state of the art.

 13

AI actors, should, based on their roles, the context, and their ability to act, apply a systematic risk management approach to each phase of the AI system lifecycle on an ongoing basis and adopt responsible business conduct to address risks related to AI systems, including, as appropriate via co-operation between different AI actors, suppliers of AI knowledge and AI resources, AI system users, and other stakeholders.

 14

Risks in paragraph 13 include those related to harmful bias, human rights including safety, security, and privacy, as well as labour and intellectual property rights.

 Schedule 2

Section 11

 Definitions used in this Act 1

“The Equality Act” means the Equality Act 2010.

 2

“Public authority” has the same meaning as in Part 1 of Schedule 19 of the Equality Act.

 3

“Algorithmic Impact Assessment” means a framework in the form laid down in Regulations made by the Secretary of State under section 4.

 4

“Algorithmic decision system” or “Automated decision system” mean any technology that either assists or replaces the judgement of human decision-makers.

 5

“Procure” means the acquisition by means of a public contract of works, supplies or services by one or more contracting authorities from economic operators chosen by those contracting authorities, whether or not the works, supplies or services are intended for a public purpose.

 6

“Test environment” means an environment containing hardware, instrumentation, simulators, software tools, and other support elements needed to conduct an assessment.

 7

“Decision support system” means an algorithmic or automated decision-making system used by a public authority intending not to base decisions solely upon its outputs.

 []

Ordered to be Printed, 9th September 2024.

© Parliamentary copyright House of Lords

This publication may be reproduced under the terms of the Open Parliament Licence, which is published at www.parliament.uk/site-information/copyright

Published by the authority of the House of Lords