The two amendments address different policy areas but share a common feature: both give Ministers broad authority to reshape online regulation through Statutory Instruments rather than primary legislation.

The first amendment – tabled during the ongoing Report Stage of the Crime and Policing Bill in the House of Lords – would insert a new power into the Online Safety Act 2023 (OSA) to enable Ministers to amend any provision of the OSA by regulations.

The only constraint on the use of the power is that it must be connected with mitigating the risks of harm arising from:

• illegal AI-generated content; or

• the use of AI services for the commission or facilitation of “priority offences” (serious offences listed in a Schedule to the Online Safety Act).

Currently, the OSA only regulates providers of AI services where those services constitute a “user-to-user service” or search engine, which does not include all AI tools. The Government has explained that it intends to use the power to bring generative AI services, such as chatbots, within the scope of the OSA framework.[4]

The amendment text explicitly states that the power “includes” the ability to apply particular OSA duties – such as preventing users encountering illegal content – to AI providers. However, as the House of Lords Delegated Powers and Regulatory Reform Committee (DPRRC) has repeatedly emphasised, saying that a power “includes” certain provisions does not confine it to them.[5]

The legal scope of a delegated power is determined by its wording, not ministerial assurances. The amendment’s text allows much more than the Government’s stated intentions.

In practice, the power could be used to:

• create new criminal offences;

• impose new fees or charges; or

• confer further legislative powers or powers to issue codes of practice or guidance.

As the Independent Reviewer of Terrorism Legislation, Jonathan Hall KC, has noted, the power could permit the creation of new offences aimed at users rather than providers, such as possessing or distributing AI tools designed to generate illegal content.[6]

While the current Government’s intention may be to target the activities of AI providers, the breadth of the power means that, once granted, it could be used by future Ministers in ways that go well beyond that intention.

The amendment provision is also a classic Henry VIII power. It would allow Ministers to amend primary legislation (the Online Safety Act) through regulations (in the form of Statutory Instruments) that generally receive less parliamentary scrutiny than a bill and cannot be amended by Parliament.

For this reason, the DPRRC has repeatedly said that such Henry VIII powers should be granted only when there is a clear and compelling justification, proportionate to the scope of the authority being conferred.[7] Beyond providing maximum legislative flexibility for Ministers, the justification is not obvious here.

The power could also be used to insert provisions into the Online Safety Act that delegate further legislative authority to Ministers. While regulations made under the top-level power would be subject to the affirmative procedure, there is no guarantee that any powers created through such sub-delegation would attract the same level of parliamentary scrutiny. For a power of this breadth, the affirmative procedure provides limited protection against inappropriate use.

Even if the Government simply intends to apply existing Online Safety Act duties to AI services, it is unclear why – other than administrative convenience – this must be done by regulation rather than primary legislation.

When the Government proposed those duties, it considered them important enough to debate and enact in primary legislation. Indeed, Parliament took a strong interest in the scope and detail of those duties on user-to-user services and search engines. It is unclear why Ministers do not think applying equivalent duties to AI providers merits the same level of scrutiny.

The second amendment relates to the Children’s Wellbeing and Schools Bill. It is a Government amendment that MPs will consider as an alternative – technically known as an “amendment in lieu” – to a change made by the House of Lords.

Peers amended the Bill to require the Government to introduce regulations obliging all social media platforms to prevent children under 16 from becoming users. The House of Commons must now respond to that amendment either by accepting it, rejecting it or proposing an alternative in its place. The Government has chosen the third option and has tabled an amendment in lieu that contains the broad delegated power that raises concern.

The Government’s proposed amendment would give Ministers the power to require platforms to prevent or restrict children under a specified age from accessing “specified internet services” or to particular features or functionalities of those services.

Crucially, Ministers would decide:

• to whom the duty to prevent or restrict access applies;

• which services or features are restricted;

• what form those restrictions take;

• the age threshold below which restrictions apply; and

• how compliance is monitored and enforced.

Regulations made under this power could also amend or repeal primary legislation, but only if such changes are to make “consequential provision”.

The scope of the power is nevertheless broad and goes beyond what the House of Lords proposed. It is not confined to social media platforms. In principle, it could apply to any “internet service” specified by Ministers, potentially including any websites, virtual private networks (VPNs), or generative AI services.

Nor is the form of restrictions set out in the Bill. Regulations could introduce age-gating, digital curfews, or limits on aggregate usage time, and these restrictions could vary between services or even between features within a single platform.

With a bill, Parliament would debate and determine choices of this kind. They are important policy decisions, not merely technical or administrative details. Yet if implemented through Statutory Instruments, Parliament would face a binary choice: approve or reject the regulations proposed by Ministers. It would have no power to amend them – for example to change which services are covered, adjust the age thresholds, strengthen or relax the restrictions, or modify the enforcement mechanisms.

The proposal must be understood in its legislative context. In deciding how to respond to the amendment proposed by the House of Lords, the Government is under considerable political pressure to demonstrate that it is taking decisive action on children’s online safety. The broad delegated power now proposed in the Government’s amendment is designed to provide a flexible response to that pressure.

The Government has justified this power on the basis that a consultation on children’s access to social media is still ongoing, meaning that detailed policy has not yet been settled. But good legislative practice suggests that if the policy is still being developed, Parliament should not be asked to legislate now. A more constitutionally coherent approach would be to complete the consultation first and then introduce legislation to implement its conclusions.

The decision to seek a broad delegated power instead reflects the political pressure on the Government to demonstrate that it is acting on children’s online safety. If Ministers want to signal their commitment, however, there are other options. The Government could undertake – or even place a statutory duty on itself – to introduce primary legislation as a priority once the consultation concludes. Instead, the Government is asking Parliament to confer a power enabling Ministers – both now and in the future – to make the substantive law in this area through regulations. While Parliament would retain a formal role in approving those regulations, it would have no power to amend them. In practice, the key policy choices would therefore be made by Ministers rather than by Parliament itself.

The Government also argues that delegated legislation is needed so it can implement the consultation’s findings “within months, not years.”[8] That contrast is overstated. A draft affirmative Statutory instrument typically becomes law within a matter of weeks, but a tightly focused bill addressing a single policy area could also pass through Parliament within a few weeks – indeed days - if necessary. Delegated legislation is typically seen within Government as a time saver but any gains in speed invariably come at the cost of significantly reduced parliamentary scrutiny and the quality of the law.

A key weakness of relying heavily on delegated legislation is its greater exposure to legal challenge. Once enacted, the provisions of Acts of Parliament are not subject to judicial review and cannot be struck down by the courts. Regulations made under delegated powers can be challenged. For example, they can be struck down for being outside the scope of the powers that Parliament has conferred, or if they are found to breach human rights obligations.

Where delegated legislation deals only with operational detail, this risk is relatively contained. But when the core elements of a policy framework are left to regulations, the entire framework becomes vulnerable to challenge. Legal proceedings – or even the threat of them – can delay implementation, undermining the speed and flexibility that the delegated powers are intended to provide.

This risk is particularly acute in a policy area such as online safety, where regulations may raise complex questions about freedom of expression, privacy, discrimination, and platform design. These are major policy choices. It is better that they are resolved through open debate and scrutiny in both Houses of Parliament than through expensive litigation in the High Court.

Even if broad delegated powers were justified in principle – for example, because it allows the Government to react quickly when unpredictable changes in technology occur – the scrutiny concerns are compounded by when these amendments were introduced. Both have been tabled late in the legislative process:

• the AI-related power was introduced during Report Stage in the House of Lords, after the Bill had already completed all its House of Commons stages; and

• the children’s internet access power appeared even later, during parliamentary “ping pong”, when the House of Commons must consider amendments made by the House of Lords.

This timing severely restricts scrutiny, particularly by MPs. Earlier stages of a bill allow for detailed clause-by-clause examination at Committee Stage, including the taking of oral and written evidence; multiple amendment opportunities at both Committee and Report Stage; and extensive time allocated for debate.

These mechanisms allow parliamentarians to raise principled and practical concerns, identify unintended consequences, refine legislative drafting, and prompt the Government to respond to criticisms or provide greater detail where provisions are skeletal.

Introducing major delegated powers at these late stages removes most of these opportunities, especially in both these cases for MPs. The stage when the Commons considers Lords amendments will be the only opportunity for MPs to influence the provisions.

Although further amendment and debate are technically possible at this stage, in practice scrutiny by MPs is constrained by strict timetabling and by limited opportunities for backbenchers and opposition parties to secure votes on amendments.

For Government bills, once the allotted time for considering Lords amendments expires, the Speaker cannot select amendments to those amendments – or amendments in lieu – for decision unless they have been moved by a Minister. Amendments tabled by backbenchers or opposition parties therefore usually fall if they are not reached before the debate ends.

The scrutiny provided at earlier stages of a bill is particularly important where the Government seeks wide-ranging delegated powers. Extended debate and meaningful opportunities for amendment allow MPs and Peers to examine the implications of such powers and to propose constraints or safeguards. These might include requiring regulations to contain specified provisions, thereby reducing the degree of policy-making discretion left for Ministers. As the Independent Reviewer of Terrorism Legislation recently observed in a note on these powers: “it is unclear how Parliament is going to be able to exercise informed legislative choice if things are done at break-neck speed.”[9]

The concerns raised by these amendments – very broad delegated powers in bills combined with limited parliamentary scrutiny of the regulations that follow – are precisely the kind of problem the Hansard Society’s proposals for reform of delegated legislation seek to address.

In a fast-moving technological and regulatory environment, governments increasingly rely on framework legislation containing broad powers so they can respond quickly to emerging issues. If Parliament had stronger mechanisms to scrutinise the regulations made under those powers, it could focus less on hypothetical concerns about how a power in a bill might be used and more on the concrete policy choices contained in the regulations themselves. In other words, the constitutional safeguard would lie not solely in tightening the drafting of the power in primary legislation, but in ensuring that Parliament – and particularly the elected House of Commons – has meaningful opportunities to examine, question and influence the Statutory Instruments that ultimately give effect to it.

A central weakness of the current system is that the scrutiny procedure for Statutory Instruments is fixed in advance by the primary legislation that confers the power. When Parliament passes a bill, it determines – often years in advance – whether regulations made under that power will follow the affirmative or negative scrutiny procedure. Yet Parliament cannot know at that stage what those future regulations will contain. This structural problem is compounded by the blunt distinction between the two procedures, which offers little flexibility to tailor scrutiny to the significance of the policy being implemented.

The Hansard Society therefore proposes replacing this rigid framework with a system in which scrutiny is calibrated according to the substance and importance of the Statutory Instrument itself. Under our model, SIs would first be sifted by a parliamentary committee and directed to different levels of scrutiny depending on their significance. Most SIs which deal with technical and administrative detail and do not require much parliamentary scrutiny would be dealt with much as they are now, if not more quickly and with less bureaucracy. But measures implementing major policy changes such as these in relation to online safety, or exercising wide delegated powers, particularly Henry VIII provisions, could receive enhanced scrutiny, including detailed examination by a new Regulatory Scrutiny Committee in the House of Commons. That Committee could take evidence from Ministers, officials and stakeholders and examine the policy and legal implications of the Instrument before it is approved.

Although the text of an SI would not be directly amendable (as that would present a number of difficulties in relation to legislative principle and practice), the motion to approve it would be amendable. MPs could therefore propose changes, setting out concerns about the Instrument and indicating conditional approval: that is, approval provided that the Government makes the specified changes. This would give MPs a meaningful opportunity to influence the content of delegated legislation, while preserving the Government’s ability to secure approval if it commands a majority.

Parliament – and the elected House in particular – currently has too few meaningful opportunities to scrutinise or influence significant policy decisions made through delegated legislation. The debates surrounding these amendments illustrate why that matters. When scrutiny of the resulting regulations is weak, Parliament is compelled to focus its concerns on the breadth of the powers in the bill itself. Reforming the scrutiny system would help resolve that tension – allowing governments the legislative flexibility they need to respond to fast-moving policy challenges, while ensuring that Parliament can properly examine, challenge and influence the regulations that ultimately shape the law.

[1] The Attorney General’s Office, Attorney General’s 2024 Bingham Lecture on the rule of law, October 2024

[2] Cabinet Office, Guide to Making Legislation, September 2025, p. 359

[3] Cabinet Office, Guide to Making Legislation, September 2025, p. 361

[4] House of Lords, Crime and Policing Bill, 2024-26, Seventh Supplementary Delegated Powers Memorandum, 2 March 2026, p. 5

[5] Examples are cited at: Vangimalla, D. (25 April 2022), Compendium of Legislative Standards for Delegating Powers in Primary Legislation (Hansard Society), p. 10, n38

[6] Hall, J., ‘Online Safety 2.0 – Some Reflections’, Independent Reviewer of Terrorism Legislation, 3 March 2026

[7] See, for example, the House of Lords Delegated Powers and Regulatory Reform Committee (2017-19), Twenty First Report, HL Paper 122, para. 11, which criticised a Henry VIII power in the Registration of Marriage Bill because it provided “for all amendments to primary legislation to be made by delegated legislation” and conferred “a very broad power, far wider than that required to achieve the policy aims…”

[8] Department for Science, Innovation and Technology, Growing up in the online world: a national consultation, CP 1528, March 2026, p. 8

[9] Hall, J., ‘Another software update to the Online Safety Act’, Independent Reviewer of Terrorism Legislation, 4 March 2026