King’s College researchers say cannabis users are more violent. Are they?

Newspapers heard a simple sentence: cannabis users are twice as likely to be violent. King’s College London had published a large review. The headline wrote itself.

The review is real. The association is real. The sentence is doing more work than the numbers can carry.

On 24 September 2026, Giulia Trotta, Victoria Rodriguez, Robin Murray, Marta Di Forti and colleagues at King’s Institute of Psychiatry, Psychology & Neuroscience published an open-access systematic review in Psychological Medicine. They pooled 63 studies covering 265,079 people. Cannabis users were more likely both to carry out violent acts and to be on the receiving end of them.

That is the finding. It is not “a joint makes you aggressive.” It is not “cannabis has nothing to do with violence.” The rest of this piece is about holding both of those facts at once.

What they actually measured

This is not one experiment in a lab. It is a stacking of other people’s work: surveys, cohort studies, psychiatric caseloads. The team searched MEDLINE, EMBASE and PsycINFO up to November 2024, then kept studies that compared cannabis users with non-users on violence — as perpetrator, as victim, or both.

Violence here is broad. They used a World Health Organization-style definition covering physical, sexual and psychological violence, in relationships and outside them. Where a study reported more than one outcome, they took the most serious. They looked separately at the general population and at people in psychiatric care. They also split self-reported violence from violence that ended in a criminal conviction.

They tried to keep other drugs out of the cannabis column. Studies that only measured cannabis mixed with alcohol or amphetamines were excluded. Residual use of other substances still sits in a lot of the original papers. The authors say so.

The number you heard, and the number underneath it

In the general population, cannabis users had about twice the odds of having perpetrated violence compared with non-users (odds ratio 2.05). Among psychiatric patients the figure was about two and a half times (OR 2.49).

Those are the numbers that became “twice as likely.” The Telegraph ran that line. King’s own press note used similar language.

Two catches, and they are not footnotes.

Most of that doubling comes from cross-sectional studies: cannabis use and violence measured at the same time. In those, the odds ratio is 2.37. Studies that follow people over time — so you can at least see which came first — shrink the association to OR 1.17. That is about 17 per cent higher odds, not a doubling. The association is still there. It is a different size of claim.

The second catch is what “twice the odds” even means. Odds are not the same as a person’s chance. When violence is rare, they sit close together. When it is common, they drift. The Conversation walked through that arithmetic after the Telegraph piece. You do not need the full lecture. You need to know that a headline “twice as likely” is a translation, and translations lose information.

Convictions look scarier still: about 3.75 times the odds in the general population, 4.21 in psychiatric samples. That is a handful of studies. Treat it as a signal, not a census.

Users were also more likely to be victims (OR 1.49). That was stronger for women than for men. Almost nobody led with that.

Are they more violent, then?

Yes, in the narrow sense the paper can support: people who use cannabis show up more often in violence statistics than people who do not, in the studies this team could pool.

No, if the question is “does cannabis turn a person violent?” The authors do not claim a single causal mechanism. They list several plausible routes — impaired inhibition, irritability in withdrawal, paranoia that makes a neutral room feel like a threat, impulsivity, the company you keep — and then say those routes sit inside a mess of other things: alcohol, other drugs, poverty, personality, the places cannabis is actually bought and used.

That last point is the one Britain should not skip. Cannabis is still illegal for most use here. Buying it can put you in rooms, with people, where violence is already more likely. Professor Simon Moore, who runs Cardiff’s Violence Research Group, told the Science Media Centre that the review does not show whether cannabis itself or those environments explain the violence. A confounder no stronger than peer group or neighbourhood, he argued, could account for the smaller follow-up figure entirely.

The paper’s own statistics do not kill that objection. Adjusted and unadjusted estimates were similar. That only means the factors those studies happened to adjust for did not move the number much. It does not mean neighbourhood, supply, and who you use with were properly measured. Often they were not.

Heterogeneity is also enormous. The studies barely agree with each other (I² in the high 80s and 90s). The pooled number is an average across different countries, ages, definitions of “used cannabis,” and definitions of “violence.” An average that noisy is a weather report, not a verdict on your Friday night.

The psychiatric finding is the one clinicians can use

The larger effect in psychiatric patients is not a surprise to anyone who has sat in a UK acute ward. King’s has spent years on cannabis and psychosis. Di Forti and Murray are not new to this argument.

In people who already have severe mental illness, continued heavy use is already tied to worse symptoms, worse adherence, more relapse. Violence risk in that group has never been “the diagnosis alone.” Substance use has always been part of the picture. This review says cannabis belongs in that picture, and that services should ask about it when they assess risk.

That is a clinical recommendation, not a moral one. It does not say a medical cannabis patient with chronic pain is a danger. It says that in a psychiatric caseload, cannabis use is common, it is often missed, and it is one of the few risk factors you can actually try to change.

Di Forti, who is a clinician as well as a co-author, put the caution in the press note herself: the review lets researchers speak more confidently about the risks “while being careful not to stigmatise people who use cannabis more broadly.”

What the paper cannot tell you

It cannot tell you about dose. Most included studies did not record how often people used, how much, or how strong it was. Subgroup analysis by THC was impossible. The authors say so.

It cannot tell you about today’s market. Most of the studies were done before high-potency flower became the default. The authors flag that too, and they suspect daily use of stronger products may show larger effects. That is a suspicion, not a result in this paper.

It cannot tell you that legalisation would raise street violence, or that it would lower it. The review is not a before-and-after of Colorado, Canada or Germany. If an illegal market is part of the association, a legal one might change the surroundings without changing the pharmacology. It also might not. This dataset cannot settle the UK reform argument, however eagerly both sides will reach for it.

It cannot tell you that cannabis is uniquely implicated. Alcohol’s relationship with violence is older, larger, and less often treated as a character test. The honest comparison is not “cannabis users versus saints.” It is cannabis versus the other drugs already in the room.

So — are they?

Cannabis users, as a group in these studies, are more often involved in violence than non-users. That includes harming other people and being harmed. The signal is stronger in psychiatric care. It gets smaller when researchers wait and watch, instead of taking a snapshot. It may be the drug, the withdrawal, the paranoia, the company, the market, or some mix. The paper cannot pick.

If you already know that a heavy, paranoid, sleep-deprived run of use can make a person worse to be around, this review is not gaslighting you. If you use at the weekend and have never thrown a punch, it is not a diagnosis. The dishonest move is the one the headlines made, and the dishonest move in the other direction is to pretend the association is a smear with no numbers behind it.

King’s researchers say cannabis users are more violent. Relative to non-users, in the studies they could pool, they show up more. That is as far as the evidence goes. It is far enough to take seriously in a clinic. It is not far enough to turn into a folk devil.

Sources

Trotta, Rodriguez, Marino, Gurmesa, Spinazzola, Li, Alameda, Di Forti, Murray and Vassos, “Cannabis use is associated with increased risk of violence: A systematic review and meta-analysis”, Psychological Medicine, 24 September 2026. King’s College London, “Cannabis use linked to higher risk of violence, major review finds”, 29 September 2026. Science Media Centre, expert reaction, 29 September 2026. The Conversation, “Cannabis doesn’t make you violent – despite what recent reports have claimed”, 15 October 2026.