62,000 Speech Arrests And No Definition Of Speech

The Home Office publishes arrest data by offence group rather than by law, which means the number of people detained under Britain's communications offences has never appeared in any official release. Big Brother Watch went and got it, force by force, under the Freedom of Information Act, and published the result on 7th September in their excellent report titled Britain's Free Speech Problem.

The headline figure was 62,199 arrests over five years. Roughly thirty-four a day, or one for every 1,113 people in the country.

The research is a real public service and the report deserves better than the coverage it got.

It also carries a handful of arithmetic slips its critics will find inside an afternoon, and a framing problem which the police records themselves expose. After those corrections, what Big Brother Watch discovered is considerably harder to rebut than what was reprinted everywhere.

Where The Figure Comes From

Big Brother Watch wrote to all 45 territorial forces in the United Kingdom, asking for arrests between 1st January 2021 and 31st December 2025 under three provisions:

1. Section 1 of the Malicious Communications Act 1988, covering indecent or grossly offensive communications sent to cause distress;
2. Section 127 of the Communications Act 2003, covering grossly offensive, indecent, obscene, or menacing messages sent over a public electronic communications network, plus persistent use of such a network to cause annoyance; and
3. Section 179 of the Online Safety Act 2023, covering knowingly false messages sent to cause non-trivial harm.

Note: the Public Order Act is also used for speech offences but does not appear to be have been included.

Thirty-eight forces returned complete arrest data.

• Police Scotland, Lancashire, and West Mercia refused on cost grounds.
• Dyfed-Powys refused for 2021 to 2023.
• West Midlands and Humberside held nothing for 2021.
• Derbyshire was still processing.

Scaling the gaps to the national average produces an estimate near 73,000.

The report's own case studies are its most powerful material:

• Ian Johnson, 68, of Frinton-on-Sea, lost his dog to an out-of-control animal, endured a campaign of harassment from the other owner, lost his temper on the telephone, and then spent two years under investigation before a court found him not guilty.
• A West Midlands schoolgirl who had merely viewed a TikTok post about a teacher had officers in her home demanding her phone under threat of arrest, and was eventually told no further action would be taken.

Those accounts do more work than any headline number could. The number ate the coverage anyway.

Corrections Worth Making

Several figures appear in more than one version, and the differences are not minor.

Three of these matter more than the rest.

1. The appendix grand total of 62,119 sits eighty short of the number the rest of the report uses, and the rows in the same table add up to the higher figure, so the headline is right and the appendix is wrong.
2. The charge total escaped in its incorrect form: the press release said "at least 18,500", the infographic said 18,510, the key findings and the charges chapter both say 18,520, and every major outlet reprinted 18,510.
3. And the national charge rate is misstated in the passage carrying the report's sharpest claim about a single force, where Thames Valley's 1.51 charges per 10,000 are measured against a "national average" of 3.4. The national charge rate is 3.1. The figure 3.4 is the national ratio of arrests to charges, a different quantity entirely, borrowed from two pages away.

The smaller slips are the ordinary consequence of shipping a hundred pages on a deadline with a small research team, and only a churl would lead on them: a statutory box heading the "Malicious Communications Act 1998", the national estimate typed as "73,00", a Lorem ipsum placeholder appearing on the map graphic, and Lincolnshire topping the arrests-per-charge table on the strength of a null value rather than a calculation.

None of this affects the 62,199 total, which is the number everything else derives from and which the appendix rows confirm.

Arrests Are Not The Same As Unique People

The key findings state 62,199 people were arrested. The arrests chapter, three pages later, states there were 62,199 arrests. The forces supplying the data are clear which is nearer the truth, and it is neither.

Cumbria Constabulary, the force at the centre of the argument, explained its own return in a disclosure of 25th September 2025. Its systems cannot link arrests to offence data, so it counts custody records by arrival date. Anyone brought into custody twice appears twice. A single custody record carrying both a Communications Act offence and a Malicious Communications Act offence is counted twice as well, once under each statute.

West Yorkshire Police warns of the same thing in almost identical language, adding a detainee may have been arrested for other offences entirely, with different outcomes.

Every force in the dataset appends a caution against comparing its figures with any other force's.

Police forces returned 62,199 communications-offence arrest entries over five years, and in at least some forces those entries are not unique people, or even unique custody events.

Sixty-two thousand entries remains an extraordinary number. It is simply not a roll-call of sixty-two thousand Britons.

Staffordshire Does Not Fit The Story

The comparison which dominated the coverage sets Cumbria's 25.69 arrests per 10,000 against Staffordshire's 0.58, and concludes Staffordshire residents are "effectively freer to speak".

Staffordshire recorded a charge rate of 3.31 per 10,000 across the same five years. The national charge rate was 3.1.

Staffordshire recorded 390 charges over the period while reporting just 68 arrest entries. These are not linked cohorts, and Big Brother Watch says as much: people are charged after voluntary interviews without ever being arrested, and the arrest and charge datasets are extracted differently.

Which is exactly why Staffordshire will not sit still inside the free-speech framing.

Its reported use of arrest runs forty-four times lower per head of population than Cumbria's, while its recorded charge rate sits slightly above the national figure. Merseyside and Northumbria show a milder version of the same pattern.

Whatever Staffordshire is doing, it is not declining to enforce these offences. It is declining to use custody to enforce them.

Cumbria remains a genuine outlier, with the highest per-capita arrest, charge, and conviction rates in England and Wales. West Yorkshire, Thames Valley, Sussex, and Northamptonshire look different again: heavy arrest, thin charge.

137 Custody Records, Four Forces

Late in 2025 an FOI requester tried a sharper approach. Rather than asking for totals, they asked forces for up to forty individual 2024 arrests, chosen by a neutral method, with the anonymised offence summary for each.

Four forces answered in full.

1. Cumbria supplied 40 records on 30th December 2025.
2. North Wales supplied 40 on 10th December.
3. Dyfed-Powys (which had refused Big Brother Watch on cost grounds) supplied 34 on 31st December.
4. Dorset supplied every eligible arrest it made in the whole of 2024: 22 arrests covering 23 offence records.

The categories below are ours, rather than any official classification, and every entry is an allegation.

Of the 122 records carrying a usable summary, five describe something an Englishman would recognise as speech in the sense the headlines implied.

The remainder reads as follows:

• A suspect who let down a victim's tyres and then telephoned to say he would come round, smash her windows, and hit her with a hammer.
• Emails and calls threatening to shoot and bomb named areas.
• A racist email followed by a screamed threat to have a group throw acid.
• Threats to slit throats, to burn down a block of flats, to throw petrol bombs at a house, and to blow up a police station.
• A Facebook post claiming the sender was wearing a bomb which would kill other people.
• Messages to a community mental health worker announcing the sender would soon afford a crossbow and three people would die (later specified as two police officers and the worker).
• Threats to stab pupils at a school.
• A man who sent threatening messages, attended the victim's address, kicked the door in, and assaulted her.
• Three separate records alleging non-fatal strangulation of a partner alongside texts threatening to kill her.
• More than two hundred calls to an ex-partner accompanied by a photograph of a knife.

The emergency-line category has no place in the free-speech argument at all, and it accounts for one record in seven.

• Over thirty calls to a control room in a single evening, with three assaults on emergency workers at the door.
• Twenty calls to police and fire claiming a partner was holding the caller at knifepoint, twice attended, twice found untrue, then continued.
• Twelve 999 calls in twelve hours after two visits and a warning.
• A caller who dialled 999 three times over lost keys, abused the control room, then telephoned the ambulance service to report an unresponsive child in the property, apparently to get officers to force the door.

All of it sits in a national spreadsheet column headed communications offences, and all of it is inside the 62,199.

Cumbria could not draw a random sample and supplied its first forty by date, every one of which falls in January 2024; North Wales, Dyfed-Powys, and Dorset span the full year, and Dorset's is a complete annual set rather than a sample.

Four forces are not a national picture.

And none of this establishes a national proportion, because no such figure exists anywhere.

It is nonetheless the only case-level evidence in the public domain, it comes from four independent forces, and it points the same way in all four.

Five Cases Which Deserve The Scrutiny

Threats are not speech. Strip out the threats, the stalking, the domestic abuse, and the 999 abuse, and five records stand out as genuine speech controversies which deserve some pressure from MPs.

1. A Cumbrian who posted repeatedly on social media accusing named people of protecting a sex offender, publishing their names and address.
2. A second Cumbrian record described in full as sending grossly offensive messages, with nothing else recorded at all.
3. A North Wales chef whose employment was terminated and who then sent WhatsApp messages his former employer considered hate-related.
4. A North Wales caller who passed private information to a neighbour implying a victim had lied about being raped.
5. A Dorset record summarised entirely as accusing a man of sex trafficking his son.

Four of the five involve publishing accusations about identifiable people, which is territory most legal systems hand to defamation and harassment rather than to criminal speech law.

Defamation (including slander and libel) is not speech.

Exactly one, the second Cumbrian entry, rests on bare offensiveness with no other conduct recorded anywhere on the file.

Those five are the civil-liberties story.

Two Thirds Of West Yorkshire's Arrests Ended In No Further Action

The strongest single piece of evidence in this affair was published in June 2024. West Yorkshire disclosed the recorded outcome of every arrest entry under section 127 and the Malicious Communications Act across two financial years, from April 2022 to March 2024.

One thousand and five of 1,427 Malicious Communications Act arrest entries, 70.4 per cent, record no further action as the highest-ranked disposal for that offence. Across both statutes, 1,289 of 1,935 entries, or 66.6 per cent, end the same way. Charge accounts for 17.8 per cent.

One arrest can appear under both statutes, so these are not 1,935 separate custody events. The disposal shown is for the communications offence, and a detainee arrested over other matters in the same episode may well have faced a different outcome on those.

What the data does not describe is the recorded outcomes of the arrests themselves, from the force's own custody system, published by the force.

Something close to two thirds of the time, in Britain's fourth-largest force, arresting somebody under these provisions led nowhere on the provision they were arrested under.

Missing: Whether Arrest Was Necessary

PACE Code G does not permit arrest on reasonable suspicion alone. An officer also needs reasonable grounds to believe arrest is necessary. Where interview is the reason for the arrest, Code G requires the officer to consider whether voluntary attendance is a practicable alternative.

Where it is, arrest for the purpose of interviewing is not necessary.

Arrest can be necessary to protect a vulnerable person; to prevent interference with witnesses or evidence; to establish identity; or to allow a prompt and effective investigation.

Given what the custody summaries actually describe, a great many of these arrests will have been necessary on one of those grounds without anybody needing to argue about free speech at all.

Big Brother Watch notes, correctly, His Majesty's Inspectorate found in April 2026 some forces now use voluntary interviews rather than arrest for suspects in online child sexual abuse cases, partly to manage resources.

The recorded necessity ground for each communications arrest is the field nobody has yet asked for, and it is the only one capable of showing whether Sussex and Staffordshire are policing different populations or applying different thresholds to the same one.

Britain Never Built Categories Of Speech

American constitutional law developed named categories of expression receiving no First Amendment protection or a reduced measure of it:

1. True threats;
2. Incitement to imminent lawless action;
3. Defamation;
4. Obscenity.

The doctrine is contested, and the Supreme Court has recently tightened the mental element required for threat prosecutions. Precisely because wrongly prosecuted threats chill protected speech.

What the categories supply is a vocabulary. An American arguing about free expression can place a death threat in a different box without conceding an inch on jokes or opinions.

Britain took the other road.

Article 10, incorporated by the Human Rights Act 1998, protects expression as a qualified right, balanced case by case against necessity and proportionality.

Article 10 – Freedom of expression

Everyone has the right to freedom of expression. This right shall include freedom to hold opinions and to receive and impart information and ideas without interference by public authority and regardless of frontiers. This article shall not prevent States from requiring the licensing of broadcasting, television or cinema enterprises.

The exercise of these freedoms, since it carries with it duties and responsibilities, may be subject to such formalities, conditions, restrictions or penalties as are prescribed by law and are necessary in a democratic society, in the interests of national security, territorial integrity or public safety, for the prevention of disorder or crime, for the protection of health or morals, for the protection of the reputation or rights of others, for preventing the disclosure of information received in confidence, or for maintaining the authority and impartiality of the judiciary.

Expression is not sorted into protected and unprotected kinds; it is weighed.

The approach has real merits (flexibility chief among them), and the ECHR is not wholly without exclusions. What it never produced is a settled category of communications which simply are not a free-speech question.

The statutes make the gap concrete.

1. Section 127 does not prohibit a kind of expression. It prohibits certain uses of a public electronic communications network.
2. Section 1 of the Malicious Communications Act reaches letters, electronic communications, or an article of any description.

Both classify by the mechanism of delivery, and by characteristics as elastic as "grossly offensive" and "menacing".

Which is why no force's spreadsheet can separate the man threatening to petrol-bomb a house from the woman who posted something tasteless, and why no FOI request will ever extract that separation.

The law never drew the line, so the data cannot show it.

"Speech arrests" is not a category in British law. It is an artefact of drafting descended from the telegraph, applied to a country of smartphones.

British judges have stated the protective half of this, and exceptionally well. In Redmond-Bate v DPP, Sedley LJ held that free speech covers the irritating, the contentious, the eccentric, the heretical, and the provocative, gloriously adding:

freedom only to speak inoffensively is not worth having.

The High Court quashed Paul Chambers' conviction in Chambers v DPP on the footing a message reasonably read as a joke in poor taste cannot be menacing. CPS guidance requires prosecutors to weigh Article 10 before charging.

All of it operates at the prosecution stage in a courtroom. None of it operates in PC Plod's 90 IQ brain at the custody desk at two in the morning.

Parliament U-Turned On Reform

In July 2021 the Law Commission published Modernising Communications Offences. It found the existing offences inconsistent, substantially overlapping, and capable of disproportionate interference with freedom of expression.

It recommended repealing both section 1 of the Malicious Communications Act and section 127(1) of the Communications Act, replacing them with a single harm-based offence pitched at serious distress.

On 4th February 2022 the Government told the House of Commons it accepted the recommendations and would repeal the existing offences.

On 28th November 2022 it changed its mind, for reasons every free-speech campaigner in the country would recognise. They bungled it. Again.

Parliamentarians and members of the public had warned the new harm-based offence could criminalise lawful and legitimate speech on the basis somebody had been "offended" by it.

The Digital Secretary removed it from the Bill. And because the replacement had gone, the Government announced it would no longer repeal the elements of the Malicious Communications Act and section 127 the replacement was meant to supersede, in order to retain protection for victims of abuse.

It's worth clarifying what happened here.

1. The Government agreed the old offences were too vague.
2. It promised Parliament they would go.
3. It then decided their intended replacement posed too great a risk to free expression, withdrew it, and consequently left the "grossly offensive" provisions standing.

This is obviously idiotic. And the executive is not required to fix this when the legislative branch (i.e. Parliament) can do it by itself.

Section 189 of the Online Safety Act 2023, in force from 31st January 2024, duly repealed only section 127(2)(a) and (b) and parts of section 1 of the Malicious Communications Act.

Section 127(1) survives.

The "grossly offensive" terminology of section 1 survives.

Two new offences were added alongside them under sections 179 and 181, each carrying requirements the old provisions lack, among them knowledge of falsity and the absence of a reasonable excuse.

Every serious institution to examine these two offences (the Law Commission and the Government among them) has concluded they are badly drafted.

They are still generating arrests at a rate of thirty-four a day.

This Isn't Simplistic Free Speech

Every arrest in this country is supposed to clear two hurdles rather than one. An officer must suspect an offence, and must also have grounds to believe taking the person into custody is necessary. When an officer makes an arrest, the reason is recorded on the custody record, in a field created for that purpose.

Nobody has ever asked for those fields, and no force publishes them.

1. We know how many arrests were made.
2. We know, in four force areas, roughly what was in the messages.
3. We do not know, in a single case anywhere in the United Kingdom, why the person was taken into a cell rather than asked to come in for an interview.

Whether the difference between Sussex and Staffordshire is a difference in populations or a difference in habits is not a mystery about human nature. It is a column in a police database nobody has requested.

Those records can be released. Cumbria, North Wales, Dyfed-Powys, and Dorset all handed over anonymised summaries of individual arrests within the cost limit when somebody asked.

North Yorkshire received the identical request, pulled its own forty cases, and then withheld every description on privacy and law enforcement grounds. Four forces found a way. One did not. Even the transparency is a lottery.

Big Brother Watch should protect the work it has done. A corrected appendix, one agreed charge figure, and the words arrest entries in place of people would cost its argument nothing and remove the only line of retreat now available to the forces it criticises. The counting was the hard part, and those numbers are already damning.

What the count cannot do is tell anyone whether a country arresting thirty-four people a day over messages is genuinely protecting victims from words which are not protected speech at all, or reaching for handcuffs out of basic tyranny.

Both are happening.

Neither the Home Office, nor the inspectorate, nor any police force in Britain publishes a figure which would tell you the proportion, and until one of them does, everybody arguing about it is guessing.