On the 9 September 2024, Lord Clement-Jones, a life member of the House of Lords, proposed the Public Authority Algorithmic and Automated Decision-Making Systems Bill [HL] “the Bill”. The Bill is currently at the second Parliamentary stage.  If enacted, the Bill would create new regulations for the use and procurement of automated and algorithmic tools in decision-making processes within the public sector.

We set out the key aspects of the Bill below, including, on when it applies, systems it prohibits, intended dispute resolution mechanisms, and how it proposes - via further regulations - to introduce Algorithmic Impact Assessments, Algorithmic Transparency Records, and obligations on monitoring, AI training, and logging.  

Whilst at an early stage in the House of Lords, it is useful to see what's on the legislative agenda on the proposals for what can (and can't) be used and how, at a time when the EU AI Act has been enacted (see our flowchart here), the UK government is developing its plans for AI regulation in the UK, and there are other AI regulatory proposals in the UK (see here).

Purpose of the Bill

There are two primary objectives of the Bill:

1. Ensuring that algorithmic and automated decision-making systems are deployed in a manner that:
   1. accounts for and mitigates risks to individuals, public authorities, groups and society as a whole, and 
   2. leads to efficient, fair, accurate, consistent, and interpretable decisions.
2. Making provision for an independent dispute resolution service.

When would the Bill apply?

The Bill applies to any algorithmic or automated decision-making system developed or procured by a public authority. That includes systems used to inform, recommend or make administrative decisions, including those in development (except if in a test environment).

However, it would not apply to automated decision-making systems used for the purpose of national security and automated systems which merely calculate and implement formulas which would automate a process of calculation which would otherwise be carried out manually and fully understood. 

Algorithmic Impact Assessments

The Bill would require a public authority to complete an Algorithmic Impact Assessment before deploying any algorithmic or automated decision-making system, subject to exceptions, such as for algorithmics used to solely formulate policy.  The Algorithmic Impact Assessment must be updated when the functionality, or the scope, of the algorithmic or automated decision-making system changes. The Algorithmic Impact Assessment would then be published in an accessible format, although it does not specify where it would be published or what is meant by accessible format.

The Secretary of State will be responsible for prescribing the form of the assessment which will have the aims of:

1. procuring, developing, and implementing algorithmic and automated decision-making systems so that the decisions made by a public authority and responsible, complying with procedural fairness and due process requirements, and its duties under the Equality Act and the Human Rights Act 1998.
2. assessing the impact of algorithms on administrative decisions, minimising negative outcomes, and evaluating the potential to maximise positive outcomes.
3. making data and information on the use of algorithmic and automated decision-making systems in public authorities available to the public.
4. understanding and reducing the risks associated with algorithmic and automated decision-making systems.
5. introducing the appropriate governance, oversight, and reporting and auditing requirements that best match the risks associated with the application envisaged.
6. undergoing responsible innovation of algorithmic and automated decision-making systems.

This framework must also include requirements for the detailed description of the algorithmic or automated decision-making system, the assessment of the relative benefits and risks of the system, what steps should be taken to minimise those risks, and an independent external security of the efficacy and accuracy of the system and a mandatory bias assessment of the decision-making system. 

Algorithmic Transparency Records

The Bill would require public authorities to give notice on a public register about either the use of or procurement of an algorithmic or automated decision-making system by way of an Algorithmic Transparency Record, subject to limited exceptions where the system is used for policy making. The record must be updated when the functionality or scope of the system changes. The Secretary of State will prescribe the form of the record, but must include:

• a detailed description of the algorithm of automated decision-making system;
• an explanation of the rationale for using the system;
• information about the system’s technical specifications;
• an explanation of how the system is used to inform administrative decisions; and
• information on human oversight of the system.

By way of comparison, the UK already has an Algorithmic Transparency Recording Standard (here) which the government has announced it intends to roll out further to additional government departments (with further details expected).

Requirements of public sector organisations on use of algorithmic or automated decision-making systems

The Bill also mandates public authorities to develop processes to monitor the outcomes of the algorithmic or automated decision-making system to help safeguard against unintentional outcomes and verify their compliance.  There are also audit and evaluation requirements for public authorities to mitigate risks that will be prescribed by the Secretary of State.

AI Training

Public authorities using such systems will be obligated to implement organisational practices and measures to ensure that employees applying final decisions have the authority and competence to challenge the system’s output.  This will mean that adequate employee training in the design, function, and risks of the system, is required to be able to review, explain and oversee its operations. 

The training must be in accordance with specified principles (Schedule 1), including:

• Inclusive growth, sustainable development and well-being
• Respect for the rule of law, human rights and democratic values, including fairness and privacy
• Transparency and explainability
• Robustness, security and safety
• Accountability

AI literacy is a hot topic and these requirements bear some similarity with other laws, proposed and enacted, on AI literacy – see our article here.

Logging

Additionally, there are design requirements for algorithmic and automated decision-making systems in the Bill which explains that there must be logging capabilities enabling the automated recording of events during operation that enable the monitoring of the operation of the system in relation to risks and legal obligations. These logs must be regularly accessible to the public authority in question and will usually to be held for a minimum period of 5 years, along with a record of whether the final decision taken followed the recommendation of the algorithmic or automated decision-making system.

Prohibitions on AI use

Public authorities will not be allowed to deploy or use an algorithmic or automated decision-making system where there are practical barriers which limit the effective assessment or monitoring of the decision-making system in relation to individual outputs or aggregate performance. Those barriers may include contractual, technical or IP issues. 

Dispute Resolution

The Bill further would require the Secretary of State to ensure that the ability to challenge decisions or obtain redress for decisions made by the algorithmic or automated decision-making system is made available through an independent dispute resolution service which is appropriate to the nature of the public authority and the decisions in question.

Next steps

Generally speaking private members’ bills in the House of Lords do not become law.  However, they are worthwhile paying attention to. Sometimes they do become law – Parliament has produced a list (here) of 405 since 1983 (in the House of Commons or House of Lords). Also, the Bill may help set galvanise and set political debate, even if on some but not all topics. Accordingly, it is useful to be aware of, and compare and contrast against other proposed legislation. It is notable that the Bill focuses specifically on the public sector procurement and use of AI – likely for the reasons we cover in our chapter on Public Law and Procurement Law and AI in the practitioner’s textbook the Law of Artificial Intelligence (see here).  Further explanation or the Bill, and indication of which parts may (or may not) gather political support, is anticipated as the Bill progresses through debates in the House of Lords.

If you would like to discuss how current or future regulations impact what you do with AI, please contact Tom Whittaker, Brian Wong, Lucy Pegler, David Varney, Martin Cook or any other member in our Technology team.

For the latest on AI law and regulation, see our blog and sign-up to our AI newsletter.

This article was written by Mopé Akinyemi​​​​.