I have never found it easy to think well of the Adam Smith Institute. I attended various of its functions during the 1980s and 1990s, when I was younger and more patient, and when I was still under the delusion that invitations to gatherings described as “libertarian” might bring me into contact with libertarians. They seldom did. The other people present were almost invariably drawn from that familiar class of grifters and corporate shills who constitute so much of what is fraudulently called the conservative and libertarian movement. I once even saw Michael Howard at one of these functions. I took care not to meet him.

The Institute’s publications have generally been worthless. Their average message can be reduced to the proposition that the State should make itself more efficient, so that it can provide us with still more of the government we are compelled to pay for. The language is of markets and liberty. The substance is an endless rearrangement of administrative furniture. A bureaucracy should be privatised—not abolished. A tax should be simplified—not repealed. A state function should be contracted out to one of the Institute’s friends—not discontinued. Instead of having our throats cut by a salaried official, we are invited to rejoice that the work has been put out to competitive tender.

It is therefore with something approaching astonishment that I find myself praising an Adam Smith Institute publication. Justice, even to such an institution, requires me to say that its Freedom of Speech Bill is a worthwhile and in several respects courageous attempt to restore a liberty that the British once took for granted and have lately allowed to be filched from them. It is not another proposal for a commission, a regulator or a stakeholder consultation. It identifies censorship as an evil. It identifies the State as the principal author of that evil. It attempts to prohibit not only direct censorship but its modern equivalents—regulatory intimidation, employer discipline, ideological conditions, secret record-keeping, compulsory affirmations and the use of supposedly independent intermediaries. That is much to its credit.

The Bill is nevertheless internally inconsistent and in places alarmingly loose. It combines several excellent principles with drafting that would keep barristers occupied for years. Some of its promises are broader than its operative protections. Some of its exceptions may swallow the rules to which they are attached. Its treatment of employment is both radical and oddly timid. Its protection against debanking is valuable but incomplete. Its treatment of internet platforms appears to surrender almost everything that its opening provisions promise. Three of its schedules are placeholders. Several important cross-references give the Bill the wrong name. In short, this is a good Bill in need of severe attention.

The following assessment is based on the complete draft as published, including its schedules. The text cited throughout is the draft found here on the 9^th September 2026.

I. What the Bill is trying to do

The long title declares that the Bill is intended:

“to recognise and restore the ancient liberty of free speech; to protect expression by the public, subject only to narrow and objective exceptions; to restrict the power of public authorities and essential services to interfere with lawful expression; to repeal or amend enactments which criminalise expression by reference to offence or distress.”

This is admirably plain. It rejects the dreary modern assumption that freedom of speech is a concession dispensed by authority after it has weighed the speaker’s interests against an indefinitely expandable catalogue of social anxieties. The liberty is described as ancient; the task is restoration rather than invention.

Clause 1 develops the purpose. The Bill is to protect discourse on:

“politics, morality, philosophy, or any other matter of public interest, including expression that others regard as offensive, insulting, abusive, shocking or hateful.”

It also intends to confine criminal liability to traditionally recognised categories such as unlawful threats, direct incitement and serious harassment. Most importantly, clause 1(2)(c) says that it will prevent suppression:

“via direct prohibitions or indirect actions which inhibit or discourage such expression.”

The reference to indirect action is essential. The characteristic instrument of present-day censorship is not the policeman dragging a pamphleteer from his bed. It is an employer receiving an informal telephone call, a bank discovering that a customer presents “reputational risk,” a professional regulator questioning whether dissent is compatible with membership, a university beginning a conduct investigation, a charity fearing for its grant, or a social-media company being privately encouraged to adjust its systems. The censor now prefers to be invisible. He issues no formal prohibition and leaves no appealable decision. He arranges consequences. The Bill understands this. Whether it defeats it is another question.

II. A welcome repudiation of the right not to be offended

Clause 5 contains perhaps the best sentence in the Bill:

“There is no right in law not to be offended by the expression of others.”

This ought to be painted above the entrance to every police station, university, employment tribunal, professional regulator, broadcasting organisation and human-resources department in the country.

Subsection (2) provides that a person does not commit an offence or incur civil liability merely because his expression is:

“offensive, grossly offensive, insulting, abusive, indecent, shocking or distressing,”

or because it has caused an emotional or intellectual impact on the hearer.

The qualification “merely because” is sensible. Speech may form part of fraud, blackmail, a threat, incitement or harassment. Its use in such conduct cannot be immunised simply by calling it expression. But offence, shock and distress are not in themselves proper grounds of legal liability. This is a direct rejection of the conceptual corruption that has overtaken British public life. Words are now routinely classified as violence, safety is redefined to include emotional tranquillity, and disagreement is redescribed as the erasure of another person’s identity. Once these metaphors are accepted, censorship follows almost automatically. If words are violence, the censor becomes a peacekeeper. If disagreement makes people unsafe, the employer who sacks a dissident imagines himself to be performing a safeguarding function. Clause 5 attacks the premise rather than merely moderating its consequences.

Clause 6 reinforces the point. No person may be subjected by the State to criminal, civil or disciplinary penalties solely because of expression on public affairs, morality, philosophy, religion or politics unless the expression belongs to one of the Bill’s unprotected categories and all conditions for liability are strictly satisfied. Again, this is good. A free society is not one in which magistrates graciously decide that an unpopular opinion was reasonable. It is one in which the opinion is none of their business.

III. The Bill’s conception of the State

Clause 2 provides that the obligations imposed on the State apply to any public authority. It then says that “public authority” includes:

“any person certain of whose functions are functions of a public nature”;

“any institution subject to direct or indirect control” by a public authority; and

“the employees, servants, and agents” of covered persons or entities.

It directs courts to construe the term expansively.

The intention is sound. Formal legal independence ought not to determine whether an institution is exercising public power. A body should not escape merely because it is incorporated as a company, registered as a charity, given a royal charter, called an independent regulator or placed at one remove from a government department. This is particularly important in Britain, where governmental functions have been distributed among a luxuriant undergrowth of agencies, contractors, trusts, universities, professional bodies, regulators, grant-funded charities and arm’s-length organisations. The citizen may experience these bodies as wielders of compulsory power even when their lawyers insist that they are technically private.

The present Human Rights Act approach to hybrid public authorities has generated difficult arguments about whether a body is performing a public function in the particular circumstances. Official discussions of the existing model recognise that a body may act in more than one capacity and be covered only in respect of its public functions. See, for example, the official explanation of the comparable public-authority test in the Environment Act 2021.

The new Bill may go further. Clause 2(2)(b) appears to cover any person “certain of whose functions” are public, without expressly restricting the Bill’s obligations to the exercise of those functions. If that is deliberate, it should be stated clearly. If a university performs public educational functions but also owns commercial property, does the Bill bind it in both capacities? If a contractor supplies a public service and operates a separate private business, is the whole company part of the State? If a charity receives public money for one programme, are all its employment decisions covered?

The present language does not answer these questions. There is also the elementary drafting defect in the expression “any person certain of whose functions are functions of a public nature.” This may be grammatically defensible after inspection, but legislation should not require inspection merely to discover its grammar. “Any person whose functions include functions of a public nature” would be cleaner.

More fundamentally, the Bill should specify the consequences of indirect control. Relevant evidence might include:

• appointment and removal of directors;
• dependence on public finance;
• statutory powers;
• ministerial directions;
• contractual control;
• regulatory compulsion;
• operational supervision;
• access to privileged governmental information; and
• the practical ability of a public authority to procure a particular decision.

Without such criteria, “indirect control” may become either a powerful anti-evasion provision or an empty phrase, depending on the disposition of the court.

IV. The prohibition of State censorship

Clause 13 is the constitutional centre of the Bill. It prohibits the State from directly or indirectly:

“prohibit[ing], restrict[ing], or penalis[ing], or threaten[ing] to prohibit, restrict, or penalise, lawful expression”;

subjecting a person to detriment in:

“employment, office, education, access to services or benefits”

because of lawful expression;

imposing ideological conditions upon:

“any licence, authorisation, recognition, funding or other benefit”; or

maintaining any policy or code of conduct that purports to penalise lawful expression.

This is excellent in conception. The inclusion of threats is especially valuable. A censor need not carry out a sanction if the threat is sufficient to produce obedience. A police officer who tells a speaker that continued lawful expression “may lead to consequences” has already chilled speech. So has a regulator who hints that a licence might be reviewed, a university that opens a disciplinary investigation without eventually imposing a penalty, or an employer that orders an employee to attend an ideological re-education session.

Clause 13(2) permits:

“neutral and reasonable requirements relating to time, place and manner of expression which are necessary.”

Some exception is required. The right to speak does not entail a right to deliver a lecture through a megaphone outside someone’s bedroom at three in the morning. But the subsection should require that a restriction be content-neutral, viewpoint-neutral and the least restrictive reasonably available means of serving a specified operational purpose. “Necessary” is a helpful word, but courts have proved capable of discovering necessity in the most convenient places.

The Bill should also say expressly that anticipated hostile reaction is not a lawful basis for restriction. Otherwise, authorities may suppress the speaker in the name of managing the audience. The old heckler’s veto is one of the easiest devices by which a purportedly neutral public-order restriction becomes viewpoint censorship.

V. Non-crime hate incidents and political dossiers

Clause 14 prohibits the State, for the purpose of punishing or chilling lawful expression, from creating, maintaining or sharing a record identifying a person as having engaged in lawful expression. It specifically prohibits recording lawful speech as a “non-crime hate incident” or under any similar name. This is another good provision, but its mental element is too restrictive. The claimant must apparently show that the record was created, maintained or shared:

“for the purpose of punishing or chilling lawful expression.”

A police force will say that its purpose was not to punish or chill. Its purpose was to monitor community tensions, assess risk, maintain intelligence, reassure complainants or identify patterns. The fact that chilling was the predictable and perhaps welcome effect will be denied.

A better rule would cover records whose purpose or reasonably foreseeable effect is to punish, stigmatise or chill lawful expression, unless the authority proves that the record is strictly necessary for the investigation of a specified suspected offence.

Clause 14(2) already allows collection and analysis where reasonably necessary for investigating crime or suspected crime. That is sufficient protection for legitimate policing. It should not be supplemented by an endlessly pliable power to create ideological intelligence files concerning people suspected of nothing illegal.

Subsection (3), requiring unlawful records to be destroyed without delay, is welcome. But it should be accompanied by duties:

• to notify affected persons;
• to identify everyone to whom the information was disclosed;
• to correct downstream records;
• to prevent the data from influencing employment, vetting or regulatory decisions; and
• to preserve a sealed copy for the limited purpose of proving the violation.

Immediate destruction without evidential preservation might otherwise deprive the victim of the material needed to establish damages.

VI. Outsourced censorship: the heart of the modern problem

Clause 15 declares that the State must not:

“arrange for, attempt to, or procure any act or omission by another person which, if done by the State, would contravene this Act”;

nor:

“fund, subsidise or otherwise support any programme or scheme whose purpose or effect is to suppress lawful expression”;

nor enter an agreement with the aim or intent of suppression.

This is among the Bill’s most important provisions. The State should not be permitted to do through a nominally independent body what it is forbidden to do directly. If a department cannot lawfully prosecute a critic, it should not be able to telephone his employer. If it cannot ban a newspaper, it should not be able to frighten its payment processor. If it cannot order the deletion of a lawful internet post, it should not be able to maintain a privileged channel through which “concerns” are transmitted to a platform.

The language “arrange for, attempt to, or procure” is broad. Yet it may still prove too narrow for the informality of modern government. Officials seldom issue an instruction saying: “Please censor this person because we dislike his lawful opinion.” They express concern. They ask whether the intermediary is aware of the matter. They mention regulatory obligations. They share intelligence. They draw attention to reputational risk. They announce a partnership against harmful content. They invite a bank, employer or platform to exercise its own independent judgment—having made clear what judgment is expected.

The Bill should therefore prohibit the State from requesting, encouraging, inducing, facilitating, coordinating, materially assisting or applying pressure calculated to produce an act that would be unlawful if performed directly. It should also provide that nominal voluntariness is immaterial. Otherwise, the State will contend that it merely supplied information and that the employer, bank or platform reached its own decision.

Clause 15(2) makes void a term in a contract, grant agreement or similar instrument where suppression is its objective and likely result. This is too demanding. A term may be dressed in the language of safety, cohesion, misinformation, values, inclusion or reputational standards. Its stated “objective” will not be suppression, though its operation may be censorious.

The test should be whether the purpose or reasonably foreseeable effect of the term is to suppress, deter or penalise lawful expression.

VII. Human-resources censorship

The Bill would often cover the use of a formally independent body’s human-resources department as an instrument of State censorship. If a government department, public regulator or other public authority asked or pressured an employer to discipline someone for lawful expression, clauses 13 and 15 would probably apply. The State would be indirectly subjecting the speaker to detriment in employment and arranging an act that it could not lawfully perform itself. If the employer were under indirect State control, clause 2 might make the employer itself a public authority. If its HR staff or consultants acted as agents of the State, clause 2(2)(d) might bring them within the same obligations. But “probably” is not good enough. Anti-evasion legislation must be explicit.

Suppose an official contacts an independent charity’s chief executive about an employee who has criticised government policy. The executive sends the communication to an external HR consultancy. The consultant conducts an investigation and recommends dismissal on the ground that the employee has damaged trust and confidence. The trustees formally approve the recommendation. Who has contravened the Bill?

The department may have procured the result. The charity may be indirectly controlled or may be acting independently. The HR consultancy may be an agent of the charity but not the State. The trustees may say they exercised independent judgment. Each participant may point at another.

Clause 21 expressly provides a cause of action against the State or an essential service provider. It does not clearly impose direct liability on every private intermediary who knowingly participates in State-procured censorship.

The Bill should provide that:

1. an act by an employee, contractor, investigator, professional adviser or HR consultant acting with actual or apparent authority is treated as the employer’s act;
2. anyone who knowingly causes, induces, assists or facilitates a prohibited detriment is directly liable;
3. the State and participating intermediaries are jointly and severally liable;
4. formal independence and the retention of nominal discretion are irrelevant; and
5. once communication and subsequent detriment are established, the burden lies on the defendants to disprove a causal connection.

Without provisions of this kind, institutional complexity may become a mechanism of immunity.

VIII. Compelled speech

Clause 16 prevents a public authority from requiring a person, as a condition of employment, education, licensing, accreditation, professional registration, funding, public benefit or citizenship, to:

“declare, adopt or affirm any political, moral, religious or ideological belief.”

It also protects refusal to attend training, ceremonies or events intended to secure such an affirmation.

This provision is both necessary and overdue. The modern ideological employer often avoids demanding a direct declaration. Instead, it requires participation in training whose assumptions are not open to dispute, adoption of preferred terminology, completion of questionnaires, pledges to advance contested political objectives, or demonstration of an approved “commitment” in applications and promotion exercises.

Clause 16 reaches some of this, especially when an event is intended to secure adoption or affirmation. But it should cover more than explicit declaration. A person may be compelled to demonstrate adherence through conduct, prescribed language or answers designed to test orthodoxy.

The exception for neutral professional or technical standards is proper. But neutrality should not be established merely because the regulator calls its standard professional. The Bill helpfully says that a standard is not neutral if compliance requires endorsement of an ideological viewpoint.

The safeguarding and security-clearance exception in clause 16(4) is more troubling. “Safeguarding” has become one of the broadest words in the administrative vocabulary. It can mean protection against real abuse; it can also mean shielding adults from lawful disagreement. The exception should apply only to measures necessary to prevent a specified and objectively evidenced risk of physical abuse, exploitation or unlawful conduct—not ideological discomfort.

IX. Employment protection: radical promise, timid execution

Schedule 3 inserts new provisions into the Employment Rights Act 1996 protecting lawful expression outside work.

A worker would have the right not to suffer detriment because of lawful expression where it was:

“in a personal capacity,”

“outside the workplace,”

and:

“otherwise than in the course of the worker’s duties.”

Dismissal for such expression would be automatically unfair, and the ordinary qualifying period would not apply. This is a major advance. A person should not ordinarily surrender his civil personality when he accepts employment. The employer purchases work, not the employee’s soul, vote, religion, friendships and opinions. Yet the protection is drawn far too narrowly. It excludes expression:

• within the workplace;
• in the course of professional duties;
• at a work-related event;
• in scholarship or teaching;
• in internal discussion;
• in criticism of institutional policy;
• in communications with colleagues;
• by a journalist as part of his work;
• by a doctor questioning medical policy;
• by a lawyer advancing an unpopular argument;
• by an academic pursuing controversial research; or
• by an employee refusing ideological terminology during working hours.

These are not peripheral cases. Many of the most serious attacks on speech occur precisely because the expression concerns the person’s professional responsibilities.

Schedule 3 then gives the employer a defence where its action was:

“a proportionate means of achieving a legitimate business aim.”

This is a large escape hatch. Almost any employer can articulate a legitimate aim: reputation, confidence, collegial relations, customer expectations, commercial partnerships or the welfare of staff. The question then becomes one of proportionality, with all the expense and uncertainty that implies. The Schedule tells the tribunal to consider whether expression materially impaired the worker’s ability to perform the job. That is useful, but it is not made a necessary condition of the defence.

The better rule would require proof of serious and material impairment of an essential function of the role. The employer should also have to prove that the impairment did not arise merely because other people objected to the speaker’s viewpoint, and that no less restrictive measure was reasonably available. The hostile reaction of activists, customers or colleagues should not be converted into a lawful reason for dismissal. Otherwise, the employer merely rewards the mob.

X. Lawful expression as a protected characteristic

Schedule 4 proposes to add “lawful expression” to the protected characteristics in the Equality Act 2010. A person would possess this characteristic if he:

“has engaged, proposes to engage, or is believed to have engaged, in lawful expression.”

This would apply to services, premises, employment and education.

The proposal is ingenious and potentially effective. It borrows an established machinery of liability and applies it to ideological discrimination. An employer, university, landlord, bank or service provider could not treat someone unfavourably because of lawful expression. It also protects intended expression and expression wrongly attributed to the claimant. This is important. An institution should not escape merely because it censored the wrong person.

There are, however, conceptual problems. The Equality Act was designed around characteristics such as race, sex, disability, age and religion. “Lawful expression” is not a stable status. It potentially includes every lawful communication, from a political manifesto to a request for tea. If interpreted literally, almost everyone shares the characteristic almost continuously.

Questions immediately arise:

• What is the proper comparator?
• What group suffers the disadvantage for indirect-discrimination purposes?
• Does unfavourable treatment have to concern the viewpoint expressed, or is any relationship with a lawful communication sufficient?
• When does criticism of expression become harassment?
• Can the speaker use the Act against a person whose contrary speech causes offence?
• How do the existing statutory exceptions apply to this new characteristic?

The Bill attempts to protect freedom from compelled expression through proposed section 12C. It rightly distinguishes refusal of a message from refusal of a person because of an existing protected characteristic. But difficult conflicts remain. The better approach may be to create a separate statutory prohibition on discrimination because of lawful expression, using the useful parts of Equality Act procedure without pretending that expression is identical in nature to every existing characteristic. At the very least, the Bill should restrict the new protection to expression concerning politics, morality, religion, philosophy, science, scholarship, public affairs and institutional misconduct. Its present definition is needlessly unlimited.

XI. Debanking

Clause 19 prohibits an essential service provider from refusing or withdrawing a core service because a person has engaged, or failed to engage, in lawful expression. This would probably prevent political debanking of the kind alleged in the cases involving Nigel Farage and George Galloway, provided that lawful expression were proved to be a material reason for the bank’s decision. Current accounts fall within the incorporated statutory definition of core services: facilities for accepting deposits, withdrawing money and making payments. These definitions appear in sections 142B and 142C of the Financial Services and Markets Act 2000, as inserted by the Financial Services (Banking Reform) Act 2013.

Clause 3 also ensures that the protected customer may be:

• an individual;
• a company;
• a charity;
• an association; or
• a political party.

The Bill therefore covers both personal and organisational accounts. The difficult issue is motive.

In a Farage-type case, the bank may rely on wealth thresholds, commercial eligibility or regulatory costs while internal documents discuss political opinions and reputational compatibility. Clause 23 would assist the claimant by shifting the burden once expression and an apparently punitive act were shown. A court could examine timing, consistency, internal evidence and whether a neutral standard was a pretext.

But clause 19 says only “on the ground that” the customer engaged in lawful expression. It should say explicitly that speech need only be a material reason or a material influence, whether or not it was the sole or principal reason. A bank should not escape by saying that ideology accounted for forty per cent of the decision and commercial considerations sixty per cent.

The permitted exceptions are sensibly narrow. A provider may act where reasonably necessary to prevent an offence, comply with a court order or comply with a specific statutory or regulatory duty. This should preserve sanctions, anti-fraud and money-laundering controls. It does not appear to permit closure merely because a customer’s lawful views create “reputational risk.” That is correct. A bank is not a political club. Its function is to hold money and process payments.

But the Bill should say so expressly:

A provider may not treat controversy, complaint, threatened boycott, adverse publicity or hostility to the content of lawful expression as reputational, integrity, conduct or relationship risk justifying withdrawal of service.

The definition of essential services is also incomplete. It may not cover every:

• electronic-money provider;
• payment processor;
• merchant acquirer;
• investment platform;
• insurer;
• crowdfunding service;
• cryptocurrency exchange; or
• provider of unregulated business credit.

A person whose bank account remains open but whose card-processing facilities are withdrawn may be just as effectively excluded from economic life.

The Bill should protect all facilities reasonably necessary to receive, hold and transmit money, make and receive payments, obtain ordinary credit, insure property and participate in lawful commerce.

The FCA has reported that its broad reviews did not find evidence in the samples examined that political views had caused account closures, while acknowledging evidential limitations and investigating reputational-risk practices. Its findings do not determine every disputed individual case, but they demonstrate the importance of documentary disclosure. See the FCA’s 2024 review of payment-account access and closures.

The Bill should require banks to state their reasons, preserve all relevant documents and identify every statutory or regulatory duty relied upon. Where financial-crime law prevents disclosure to the customer, the full records should still be available to a court under confidential procedures.

XII. Internet services: an extraordinary surrender

Clause 20 is the most disappointing provision in the Bill. It provides that no provider or user of an interactive computer service shall be treated as the publisher of content supplied by another provider. Some protection against publisher liability is desirable. A platform that merely carries other people’s speech should not be treated as though it wrote every sentence.

But clause 20(2) then immunises action taken to restrict material the provider considers:

“harmful, obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable.”

The words “otherwise objectionable” could scarcely be broader. They allow a platform to suppress lawful political expression for almost any reason, provided the platform itself objects to it. Thus the Bill solemnly proclaims that offence, insult and shock are no grounds for State censorship, then protects the internet intermediary that suppresses expression because it considers the material objectionable.

The result is a peculiar division:

• the State may not censor lawful political opinion;
• the dominant platforms may censor it freely;
• and they receive statutory immunity for doing so.

Clause 15 might apply if the State arranged or procured the platform’s decision. But where the platform acts on its own, or where governmental influence cannot be proved, the speaker has no remedy. This may reflect a libertarian belief that private property settles the matter. That view becomes less convincing when a handful of platforms control the practical means by which political communication reaches the public, when those platforms benefit from statutory immunities, and when their decisions are entangled with advertising markets, financial services, app stores and regulatory pressures.

One need not require platforms to carry everything. But any platform claiming statutory immunity from publisher liability should owe corresponding duties:

• to apply published rules consistently;
• to distinguish unlawful content from merely unpopular opinion;
• to give reasons;
• to provide notice and appeal;
• not to discriminate by political viewpoint; and
• not to coordinate secretly with public authorities.

Immunity without neutrality, transparency or due process is not a free market. It is delegated sovereignty.

XIII. The definition of unlawful expression may consume the Bill

Clause 8 says:

“Expression is protected by this Act unless it falls within a category of unlawful expression specified in this Act.”

This suggests an exhaustive list. The Bill then identifies perjury, obstruction of justice, unlawful threats, fraud, blackmail, criminal conspiracy, direct incitement, defamation, harassment and certain national-security offences.

So far, so sensible. But clause 11(1)(l) includes:

“any other category of expression which has been expressly designated as unlawful, or in respect of which penalties or other liability may be imposed, by an Act of Parliament.”

This threatens to swallow the scheme.

The Bill’s stated ambition is to restrict unprotected speech to “narrow and objective exceptions.” Yet clause 11(1)(l) appears to preserve every statutory restriction that is not specifically repealed. If another Act penalises expression on broad grounds, that expression becomes unlawful under this Bill simply because the other Act says so.

Clause 25 adds that the Bill is to prevail over implicitly inconsistent enactments. But clause 11(1)(l) expressly imports statutory prohibitions. A court may therefore ask:

• Is the older restriction displaced by clause 25?
• Is it preserved by clause 11(1)(l)?
• Must it be read narrowly?
• Is it an unprotected category merely because Parliament once attached liability to it?

The answer is obscure. If Parliament intends the Bill’s exceptions to be exhaustive, clause 11(1)(l) should be deleted or confined to statutes expressly listed in a schedule. If Parliament intends all existing statutory restrictions to survive unless expressly repealed, the Bill does not restore freedom of speech nearly as comprehensively as advertised.

XIV. Harassment

The proposed amendment to the Protection from Harassment Act 1997 would require a course of conduct:

• on at least two occasions;
• intended to cause serious alarm or distress; and
• serving no legitimate purpose.

It also sensibly provides that communications are not intentionally directed to someone merely because that person chooses to navigate to, follow, subscribe to or monitor them. This is a valuable correction. A person should not be able to manufacture harassment by repeatedly seeking out speech addressed to the public and then complaining that he has encountered it.

But “serves no legitimate purpose” may present difficulty. Courts should not decide whether a political, religious, artistic or personal communication has sufficient social value. The proper question is whether the communication was targeted at the complainant with the intention of causing serious alarm or distress, rather than whether a judge approves its purpose.

The Bill should also distinguish:

• repeated unwanted direct communication;
• public criticism of an identifiable person;
• reporting alleged misconduct;
• satire and ridicule;
• political campaigning; and
• attempts to obtain answers from a public office-holder.

Public criticism may predictably distress its subject without becoming harassment.

XV. Direct incitement and threats

Clause 9’s test for direct incitement is modelled on a demanding and speech-protective approach. The expression must:

• be directed to identifiable persons or a specific audience;
• intend imminent lawless action; and
• be likely, in context, to produce imminent lawless action.

It adds that general advocacy or praise of unlawful conduct is not enough.

This is one of the Bill’s clearest achievements. It prevents the criminalisation of abstract advocacy, historical discussion, rhetorical excess and remote speculation. The threat provisions are also generally directed towards genuine threats of death, serious harm or unlawful violence. However, Schedule 2 contains areas requiring technical review. For example, the offence concerning fear or provocation of violence should be carefully reconciled with the Bill’s insistence that liability attach to threats and conduct rather than objectionable content.

The Bill must ensure that a speaker is not punished because opponents threaten violence in response. Schedule 2 properly says that a person may not be arrested or convicted for breach of the peace because of lawful expression alone, even where another person’s unlawful reaction is predictable. That principle should apply throughout the Bill.

XVI. Strategic litigation

Clause 22 supplies a substantial anti-SLAPP mechanism. A defendant may seek early dismissal of litigation connected with publication on a matter of public interest. The claimant must demonstrate substantial merit, likely success, a non-censorious purpose and sufficiently serious harm. Proceedings are stayed during the application. A successful defendant receives full reasonable costs, save in exceptional circumstances, and may receive damages for loss caused by the abusive claim. These are strong provisions.

Yet the application is apparently available only before service of the defendant’s answer. That may be too short and procedurally awkward. The censorious nature of a claim may emerge through disclosure, amendments, settlement demands or litigation conduct. The court should be able to dismiss a SLAPP at any stage.

The requirement that the claimant prove a non-censorious purpose may also encourage artificial disputes about subjective motive. The objective effect of the proceedings should be sufficient where the claim is disproportionate and lacks a realistic prospect of success.

The Bill should additionally protect:

• pre-action correspondence;
• threats of litigation;
• data-protection claims used to suppress reporting;
• privacy and confidence claims;
• multiple proceedings against connected defendants; and
• overseas judgments or proceedings intended to evade the domestic protection.

XVII. Remedies and costs

Clause 21 gives courts a broad remedial power, including declarations, injunctions, emergency relief, quashing orders, reconsideration orders, damages and reasonable costs.

Clause 21(4) provides that a claimant successful in whole or in part must ordinarily receive reasonable costs. This is important. A theoretical right that costs more to enforce than the claimant can afford is not much of a right.

Clause 23’s burden shift is likewise useful. The court must examine temporal proximity, inconsistent treatment, post hoc explanations, circumstantial evidence and supposedly neutral standards that penalise expression in practice.

The remedies nevertheless need expansion. The Bill should expressly authorise:

• restoration of employment or services;
• correction and deletion of blacklists;
• disclosure of communications among public authorities and intermediaries;
• protection against retaliation;
• civil penalties for deliberate concealment;
• personal liability for officials acting knowingly or recklessly;
• aggravated damages for malicious conduct;
• representative proceedings; and
• awards for distress and loss of opportunity even where precise financial loss is difficult to quantify.

Time limits also require attention. Employment claims are generally subject to short tribunal deadlines. A victim may spend months pursuing an internal grievance before discovering governmental involvement. The limitation period should run from the date on which the claimant knew, or could reasonably have known, the material facts.

XVIII. Constitutional overreach

Clause 25 says that the Act is to prevail over any implicitly inconsistent enactment, whenever passed, unless the other enactment expressly disapplies or amends it and identifies the affected provisions. This is a sensible instruction against accidental erosion. It requires a future government wishing to restore censorship to admit what it is doing. But the clause cannot permanently bind a future Parliament. Under orthodox parliamentary sovereignty, a later Parliament may legislate inconsistently. Courts may give considerable weight to the express-disapplication requirement, but the present Parliament cannot guarantee that its preferred form will control every future enactment.

The clause is therefore valuable as a rule of construction and a political tripwire, not as an entrenched constitutional settlement.

Clause 26 attempts to prevent the European Convention on Human Rights, judgments of the European Court of Human Rights and foreign law from being used to justify restrictions broader than the Bill permits. That is broadly consistent with the Bill’s purpose. Yet the drafting should distinguish between using foreign or international law to impose additional restrictions and considering it for persuasive, historical or comparative purposes. The present prohibition is unnecessarily categorical. Courts should be free to learn from other systems while remaining forbidden to dilute the domestic right.

XIX. Repeals and their unintended consequences

Schedule 1 proposes extensive repeal:

• section 127 of the Communications Act 2003;
• sections 1 to 7 of the Contempt of Court Act 1981;
• the Hate Crime and Public Order (Scotland) Act 2021;
• Article 10(2) in Schedule 1 to the Human Rights Act 1998;
• the Malicious Communications Act 1988;
• the Obscene Publications Act 1959;
• the Online Safety Act 2023;
• the Public Order Acts 1986 and 2023;
• part of the Terrorism Act 2000; and
• several provisions of the Terrorism Act 2006.

Some of these repeals are central to the restoration promised. Laws criminalising communications because they are “grossly offensive” are plainly incompatible with the Bill’s principles. But wholesale repeal followed by partial recreation demands meticulous consequential drafting. The entire Public Order Act 1986 cannot be removed safely unless every necessary non-speech offence, definition, police power and cross-reference elsewhere in the statute book is dealt with.

The Bill recreates riot, violent disorder, affray, unlawful assembly and certain threatening offences in Schedule 2. But it must be checked against every enactment relying on the repealed Act.

The proposed repeal of sections 1 to 7 of the Contempt of Court Act 1981 is especially curious because clause 11 preserves contempt of court as unlawful expression. Would repeal revive or expand common-law contempt? Which statutory protections and limitations would disappear? The Bill should reform contempt precisely, not create uncertainty over the relationship between statutory and common-law powers.

Similarly, the Bill preserves offences under the Official Secrets Act 1989 while Schedule 1 repeals section 5 of that Act. The exact intended boundary needs clarification.

A freedom-of-speech statute is no place for legislative improvisation. Broad repeals may be justified, but they require a complete consequential schedule.

XX. An unfinished United Kingdom Bill

Clause 31 declares that the Act extends throughout the United Kingdom, subject to modifications in Schedules 5 and 6. Schedule 5, concerning Scotland, says only:

“[PLACEHOLDER]”

Schedule 6, concerning Northern Ireland, says the same.

Schedule 7, the promised non-exhaustive list of public authorities, is also a placeholder.

These are not decorative omissions. Scotland and Northern Ireland possess different criminal-law, court, policing, equality and institutional arrangements. The Bill also proposes to repeal a Scottish hate-crime statute. It cannot responsibly be presented as a United Kingdom settlement until those differences have been addressed.

Schedule 7 is especially important to the anti-outsourcing purpose. It should identify:

• government departments;
• local authorities;
• police forces;
• prosecuting authorities;
• regulators;
• publicly funded universities;
• NHS bodies;
• professional bodies exercising statutory powers;
• arm’s-length agencies;
• publicly controlled companies;
• contractors performing specified public functions; and
• bodies whose funding or governance makes them susceptible to governmental direction.

The list may remain non-exhaustive, but its contents would give courts and citizens valuable guidance.

XXI. The Bill cannot remember its own name

Clause 31 says:

“This Act may be cited as the Freedom of Speech Act 2026.”

Schedules 1, 3 and 4 repeatedly refer instead to the:

“Free Speech Act 2026.”

This is not merely an aesthetic lapse. The schedules use the wrong title when incorporating the definition of “lawful expression” into the Human Rights Act, Employment Rights Act and Equality Act. A court would probably treat the error as obvious. But the central employment and discrimination protections should not depend upon a judge deciding to repair an incorrect statutory reference. For a Bill drafted to restore an ancient constitutional liberty, remembering its own title does not seem an excessive demand.

XXII. What should be done

The Bill should be revised around several principles.

First, its categories of unprotected expression should be genuinely exhaustive. Clause 11(1)(l) should be deleted or limited to expressly listed statutory offences.

Second, indirect State action should include requests, encouragement, facilitation, coordination, information-sharing and regulatory pressure—not merely formal procurement.

Third, participating intermediaries should incur direct and joint liability.

Fourth, employment protection should extend beyond speech made outside work. Restrictions should be permitted only where expression seriously and demonstrably impairs an essential function of the role.

Fifth, hostile public reaction and reputational discomfort should not constitute legitimate grounds for employment discipline, debanking or denial of service.

Sixth, financial protection should extend to payment processors, electronic-money institutions, merchant services, investment platforms and other facilities necessary for participation in economic life.

Seventh, online-platform immunity should be conditional upon viewpoint neutrality, transparency, consistent rules and meaningful appeal.

Eighth, the Bill should impose reason-giving, record-preservation and disclosure duties upon institutions accused of censorship.

Ninth, the Scottish, Northern Irish and public-authority schedules must be completed.

Tenth, the repeals require a comprehensive technical and consequential review.

Finally, the Bill should decide what it means by freedom of speech. At times it means protection only against the State. At other times it reaches employers, banks, universities and service providers. Clause 20 then retreats into an expansive defence of private platform censorship.

There is a coherent position available: the law should prevent the State and institutions controlling essential avenues of civic and economic participation from punishing lawful expression, while preserving genuine editorial choice, private association and property. But the boundaries must be stated rather than left to emerge accidentally from a collection of borrowed legal devices.

Conclusion

For all its faults, the Freedom of Speech Bill deserves serious attention. It recognises that Britain’s present problem is not confined to criminal prosecution. Censorship now operates through administrative records, employment discipline, compelled ideology, professional regulation, banking, funding, contracts and private intermediaries. The Bill attempts to deal with all of these.

Its declaration that there is no legal right not to be offended is excellent. Its prohibition of direct and indirect State censorship is strong. Its attack on non-crime hate records is overdue. Its outsourcing provision identifies a central modern abuse. Its employment and Equality Act amendments could make ideological dismissal substantially more difficult. Its debanking clause would provide a real remedy where lawful political expression materially caused the withdrawal of ordinary financial services. Its anti-SLAPP provisions are more serious than the timid reforms usually offered under that name.

But the draft promises more than it presently delivers. The preservation of every statutory category of unlawful expression threatens its supposedly narrow exceptions. Employment protection is confined too heavily to private life. The legitimate-business-aim defence may restore censorship by another route. Outsourced HR decisions are not attributed clearly enough. Financial services are incompletely defined. Platform censorship is effectively immunised. Three schedules are empty. Major repeals lack the necessary consequential architecture. And the Act is called by two different names. These are not reasons to abandon the project. They are reasons to improve it.

The Adam Smith Institute has, in this instance, done something better than call for a more efficient mechanism of government. It has proposed that government—and the network of dependent bodies through which it increasingly acts—should be deprived of powers it ought never to have possessed.

I doubt I shall ever be invited to another of its events. If I were to be, I might, for the first time in two decades, seriously consider accepting.

I might.