Wales has done something unusual – it has enacted a duty to prohibit electoral lies before deciding what the prohibition will look like. Since 27 June 2026, section 13 of the Government of Wales Act 2006 has required the Welsh Ministers to make provision against false or misleading statements of fact made to affect who is elected to the Senedd, but it leaves the decisive choices to a future Conduct Order. I argue that Wales should resist a single general offence. It should begin by separating three kinds of conduct that are more amenable to legal regulation than the rest – falsehoods about voting mechanics, demonstrably false factual claims in controlled campaign material, and synthetic impersonation. The distinction matters because each gives an adjudicator a way to identify falsity without becoming an arbiter of political truth.

The Ban Behind the Ban

The duty comes from section 24 of the Senedd Cymru (Member Accountability and Elections) Act 2026, passed by the Senedd on 17 March 2026 and in force since 27 June. The duty binds, but the prohibition will exist only when the Welsh Ministers make a new Conduct Order to succeed the Senedd Cymru (Representation of the People) Order 2025 and the Senedd approves it by resolution. The Order may define what counts as a statement of fact, cover misleading statements or only false ones, choose subjects and speakers, require knowledge or recklessness, create exceptions and, as the Act confirms, create criminal offences.

Wales is not starting from nothing. The Conduct Order already applies section 106 of the Representation of the People Act 1983. Its principal limb prohibits false statements of fact about a candidate’s personal character or conduct, whose narrow reach was confirmed in R (Woolas) v Parliamentary Election Court [2010] EWHC 3169 (Admin); subsection 106(5) separately covers falsely stating, for electoral advantage, that a candidate has withdrawn. What the 2026 Act demands is broader – provision against false or misleading statements of fact in general. That is a step no part of the United Kingdom has yet taken. In Johnson v Westminster Magistrates’ Court [2019] EWHC 1709 (Admin), at [36], the Divisional Court observed that Parliament had twice chosen not to legislate that broadly.

The politics add urgency. Plaid Cymru, the party that pressed for the ban, won the May 2026 election with 43 of 96 seats and governs in minority; Reform UK, from whose benches came the only vote against the Bill, is now the main opposition with 34 seats. The Order is expected before the 2030 election and must pass a chamber the government does not control.

The debate so far has concentrated on the legislative vehicle. The Senedd’s Legislation, Justice and Constitution Committee objected that a criminal offence should not be created by secondary legislation; the committee that scrutinised the Bill called the approach inappropriate; and Sarah Nason has set out the constitutional and Convention risks of leaving the definition to ministers. Those objections make the substantive boundary especially important. A broadly worded offence created through secondary legislation would leave unusually consequential questions about political falsity to later interpretation. The threshold question is therefore which forms of deception can be defined precisely enough for the law to prohibit them.

Different Forms of Deception

Electoral deception does not present a single legal problem. A message giving the wrong polling date can be checked against an official record. A fabricated allegation that a candidate has a conviction may also be verifiable, but it engages reputation and the heightened protection of campaign criticism. A claim about the economic effects of a manifesto rests on contested assumptions, and a broken promise is rarely provably false when made. A synthetic video raises the narrower question of whether the candidate said or did what the recording depicts. Each may influence a vote, yet the evidential task differs. A single prohibition on “false or misleading statements” leaves those differences to be worked out at the moment of greatest political pressure.

Examination of comparable jurisdictions reveals variation in what relevant laws prohibit and the fault required to establish an offence. Those laws also draw the line between claims that can be determined legally and those that cannot.

Canada is useful primarily on the issue of fault. Section 91 of the Canada Elections Act covers a closed list of false claims about candidates and associated public figures, including criminality, citizenship, birthplace, education, qualifications and group membership. Parliament removed the word “knowingly” in 2018, but in Canadian Constitution Foundation v Canada the Ontario Superior Court held that intending to affect an election is not the same as knowing that a statement is false; without a knowledge requirement, the provision failed minimal impairment under the Charter. Parliament subsequently restored the knowledge requirement. Wales need not copy Canada’s closed list. The more useful lesson concerns fault. Falsity and culpability are separate questions, and penal liability becomes harder to justify when the law collapses them.

South Australia helps with scope. Section 113 of the Electoral Act 1985 applies to electoral advertisements containing a purported statement of fact that is inaccurate and misleading to a material extent; opinion and prediction fall outside the provision. The Electoral Commissioner may request withdrawal or retraction. At the 2022 election, the Electoral Commission of South Australia received 117 misleading-advertising allegations. Of these, 83 were found not to breach the Act and ten could not be assessed because insufficient evidence was provided; only 24 resulted in regulatory intervention, including nine retraction requests. Together with ECSA’s account of common misconceptions about section 113, that gap suggests that complainants sometimes understood its reach more broadly than the statutory test allows. It also shows the value and cost of confining regulation to controlled campaign material and purported statements of fact. Even there, the Commissioner must isolate the exact proposition, distinguish fact from evaluative judgment and decide whether any inaccuracy is materially misleading.

A Threshold for Legal Intervention

France takes a different approach by setting a demanding threshold for intervention. Article L. 163-2 of the Electoral Code permits a judge to act within forty-eight hours during the three months before a national election where inaccurate or misleading allegations of fact are disseminated online deliberately, artificially or automatically and on a massive scale. In its 2018 decision upholding the law, the Conseil Constitutionnel required both the falsity and the risk to the integrity of the vote to be manifest. That limitation matters beyond the French procedure. It turns uncertainty into a legal stopping point. Where falsity depends on a contested political inference or cannot be established to a demanding standard, the law should not require an adjudicator to manufacture certainty.

Taken together, the comparators point toward a modest principle. Regulation is easiest to defend where the proposition is externally verifiable, the regulated speaker or material is defined, and fault is specified. It becomes more dangerous as the law moves from factual attribution toward causal claims, predictions and political evaluation. That distinction provides a workable basis for the Welsh Order.

Three Categories

I suggest that there should be three categories of offences under the Order. The first category should cover falsehoods about voting mechanics – dates, places, eligibility, required documents, methods of voting and counting, and the authority of election officials. These propositions are tied to an official legal or administrative record. A rule can therefore define the prohibited conduct without asking whether a campaign argument is substantively sound. Where the prohibition carries penal liability, knowledge or recklessness should be required.

The second category should cover demonstrably false factual propositions in paid advertising or in material controlled by a candidate, party, campaign or registered third party. Coverage should turn on whether the overall message communicates a separable factual proposition capable of proof or disproof. Political importance or reputational harm alone should not be enough. Opinion, satire, predictions and promises should remain outside the prohibition unless they contain a factual assertion that can be tested independently. The distinction between demonstrable facts and evaluative judgment also appears in the Article 10 ECHR case law. In Lingens v Austria (1986) 8 EHRR 407, at [46], the European Court of Human Rights held that the truth of value judgments, unlike statements of fact, is not susceptible to proof.

If the Order reaches “misleading” statements, the concept should operate through the overall-message test and only where the message communicates a materially false factual proposition capable of proof or disproof. It should not become a freestanding standard for adjudicating political argument. Materiality matters here too. Trivial inaccuracies should not become election offences.

The third category should address synthetic impersonation, where the central question is provenance rather than the truth of a policy claim. Singapore’s Elections (Integrity of Online Advertising) (Amendment) Act 2024 targets online election advertising that realistically depicts a candidate saying or doing something the candidate did not. A Welsh prohibition could likewise focus on the intentional creation, commissioning or knowing dissemination of realistic synthetic material that falsely attributes words or conduct to a candidate and is likely to deceive electors. Evident satire and clearly disclosed synthetic content should fall outside. The adjudicator would decide whether the attribution was false. The political view expressed through it would remain outside the inquiry.

The Competence Problem for Synthetic Impersonation

One constitutional constraint poses particular problems for the third category. Section 13 permits only provisions that would be within the Senedd’s legislative competence, and the Counsel General acknowledged during scrutiny that the Bill itself sat at the limit of that competence. Senedd elections are devolved, but telecommunications and internet services are reserved under Section C9 of Schedule 7A. The Online Safety Act 2023 also places significant platform-regulatory functions at UK level through Ofcom. Synthetic impersonation is therefore difficult for a Welsh instrument. The false attribution is electoral conduct, while much of the infrastructure through which it spreads is regulated at UK level.

That boundary should shape the offence itself. The Conduct Order can focus on conduct by candidates, parties, agents and registered third parties in Senedd campaigns – creation, commissioning, publication or knowing dissemination – rather than attempting to impose duties on hosting platforms. This will not capture every synthetic deception. Unaffiliated speakers may create material outside campaign-controlled accounts. Even so, the Welsh prohibition can remain tied to devolved electoral conduct without assuming powers over platform regulation that the Senedd does not possess.

A Limited Constitutional Ambition

Existing UK public law gives good reason for restraint. In R (Wheeler) v Office of the Prime Minister [2008] EWHC 1409 (Admin), the Divisional Court treated a promise to hold a referendum as lying so deep in the “macro-political” field that it could not generate an enforceable legitimate expectation, and there were no judicial standards by which to decide whether two treaties were materially equivalent. The case illustrates the difficulty of turning political promises into questions for legal determination. In Johnson, arising from Marcus Ball’s attempted private prosecution over the £350 million Brexit campaign claim, the Divisional Court refused to extend misconduct in public office to allegedly misleading campaign statements. At [38], drawing on R v Rimmington [2005] UKHL 63, the Court also stressed that a criminal prohibition must be sufficiently clear and certain for a person to know in advance what conduct is forbidden. That concern reinforces the need to precisely define the Welsh categories before enforcement begins.

The Welsh scheme should therefore leave many campaign lies beyond legal control, and that limit is right. Disputed causal claims, selective statistics, evasive promises and most political hypocrisy belong to voters, journalists, opponents and fact-checkers. Public intervention is easier to justify where the claim is verifiable, the legal category is defined in advance and the decision-maker does not have to choose among competing political interpretations.

Wales has not yet chosen between a symbolic ban, an unusably broad offence, and a workable prohibition. The Conduct Order is a rare opportunity to draw the boundary before litigation hardens it. Given the arithmetic of the seventh Senedd, a design that can be defended as narrow may also be the only one that passes. Its ambition should be limited and concrete. It should identify the parts of electoral deception that institutions can adjudicate with legal standards, require fault where punishment is serious, and leave the remainder where democratic disagreement normally belongs.

Julio Alberto Barajas Servin is Head of the Department for Coordination with Local Electoral Bodies at Mexico’s National Electoral Institute (INE) and a member of Mexico’s National Professional Electoral Service. He holds a doctorate from the National Institute of Public Administration (INAP). He writes in a personal capacity.