Has the Cold War Ceiling Become the New Floor?

The Cold War Benchmark, Now Exceeded

In the autumn of 1983, a KGB officer stationed in London filed more intelligence reports in a single quarter than his entire residency would manage in the following decade. That is the benchmark most Western analysts reached for when they needed a shorthand for peak Soviet pressure. On October 13, 2026, Sir Ken McCallum made it redundant.

The MI5 Director General's annual threat update was not a seasonal warning dressed in dramatic language. It was a structural statement: the service now allocates more resources to countering state-sponsored threats than at any point during the Cold War. The number requires a mechanism to make sense of it. The mechanism is this - the caseload for state-threat investigations has risen 400 percent over the last decade, compounding steadily while the public debate remained fixed on jihadist networks and lone-actor radicalization.

The Cold War analogy is tempting, and here is exactly where it breaks. The Soviet threat was concentrated, hierarchical, and largely containable within the category of espionage. What McCallum is describing is a different geometry: multiple state actors operating simultaneously across different methodologies, legal thresholds, and target sets. The resource pressure does not come from one adversary running hard; it comes from several running at once.

The sharpest single figure in his address was not the headline comparison but the quieter one beneath it. State-threat and counter-terrorism missions now occupy an equal share of MI5 operational capacity. For a service whose entire post-2001 institutional culture was built around the primacy of counter-terrorism, that arithmetic represents a genuine reorganization of purpose. The Cold War peaked and receded. What McCallum described on October 13 is not a peak - it is the new floor.

Nineteen Plots Disrupted, One in Five Suspects Under Seventeen

Since early 2020, MI5 and police have dismantled nineteen late-stage terrorist plots. That number is the counter-terrorism baseline, the floor of operational effort the service must sustain regardless of what else is placed on its desk. In April 2026, the UK threat level was raised to "severe" - the official vocabulary for "highly likely" - and it has stayed there.

The figures hold until you look at who is being arrested. One in five terrorism-related arrests in the UK now involves a person under seventeen. Read that as a rate, not a handful of cases. It describes the age profile of a recruitment pipeline, not an anomaly at its edge.

The structural interpretation is the harder one. Radicalization reaching children at scale indicates that the vectors - online platforms, encrypted messaging, algorithmically served content - are operating faster than any disruption programme can run in the opposite direction. The nineteen disrupted plots represent success. The age of the arrested represents a prior failure, one that precedes by months or years any intelligence intervention.

MI5 must now maintain this counter-terrorism tempo while simultaneously matching it on the state threats side. McCallum's formulation was precise: "our terrorism and state threats operational missions are pretty much of equal scale." That sentence describes a doubling of the cognitive and institutional burden, not a rebalancing. The nineteen plots since 2020 are not historical - several of the individuals arrested in connection with them were still attending school when they were detained. The pipeline feeds itself. Somewhere a case file was quietly opened on someone who will not be seventeen for another year.

The Expulsion Dividend: How Russia Lost Its Professionals and Found Its Proxies

Close to 1,000 Russian intelligence officers have been expelled from European postings since the invasion of Ukraine - the largest mass eviction of declared and undeclared intelligence personnel since the 1971 expulsions that stripped the KGB's London residency to a skeleton. McCallum's own assessment acknowledged the consequence plainly: "With its traditional covert footprint eviscerated by almost 1,000 expulsions since the invasion, the Kremlin's global toolkit is weak." That is the honest accounting. The problem is that weak and deniable is a combination with operational uses.

The Kremlin's response has been structural, not cosmetic. Without professional officers to run cut-outs, recruit assets and manage tradecraft, Moscow has turned to what McCallum called "criminals scrambled to do their bidding" - recruited via encrypted apps, paid in cash equivalents, and tasked with arson, sabotage and physical harassment on UK soil. The model is familiar from other degraded networks: when the professional tier is removed, the work gets outsourced to people who are cheaper, louder and harder to manage.

That last quality is the point. Degraded tradecraft raises the attribution noise floor. A criminal who burns a warehouse in Birmingham for a fee leaves a messier evidentiary trail than a trained officer running a blackmail operation, and the chain of command is easier to deny. The operational logic is not that proxies are better; it is that their failures are less legible. Ask the small question first: who absorbs the cost when an arson-for-hire goes wrong? Not the handler in Moscow. The proxies are, in this sense, the load-bearing wall now - and that is a structural fact, not a reassurance.

MI5's Iran Warning: Encrypted Instructions, Local Execution, RAF Fairford

The pattern is not new, only the geography. A foreign intelligence service identifies a target community, recruits a local intermediary who will never meet a handler face to face, and issues instructions through a messaging application that logs nothing and expires on schedule. What changed, in the Iranian case, is the ambition: the IRGC ran both tracks simultaneously, directing harassment and surveillance against UK Jewish communities while probing military infrastructure, the two lines of effort managed from the same operational architecture.

Over the past twelve months, more than fifty arrests in the United Kingdom were linked to Iran-connected operations. That is the largest single-country enforcement surge in recent counter-intelligence record. The number is not a sign of containment; it is a sign of volume.

The most significant disclosed case involves RAF Fairford, the Gloucestershire base that routinely receives US strategic bombers. Andy Burnham confirmed the plot was almost certainly directed by Iran's Islamic Revolutionary Guard Corps. The exact tactical nature of the operation remains undisclosed, which is itself a piece of information - the gap where a denial would normally sit.

What made prosecution possible at this scale was not a change in method but a change in law. The National Security Act 2023 replaced legislation that was built for a world of dead drops and diplomatic cover, and supplied prosecutors with instruments suited to the current architecture: encrypted coordination, deniable proxies, targets chosen for symbolic rather than strictly military value. Dozens of arrests and prosecutions have since been brought under that framework.

The IRGC is not the first service to discover that a domestic legal vacuum is an operational asset. It will not be the last. The Act closed the gap; whether the resourcing matches the caseload is the question the fifty arrests leave open.

China's Patience: Fake Profiles and University Laboratories

Russia scrambles. China waits. That distinction, understated in McCallum's briefing, is the structural difference the headline numbers obscure.

Where Moscow now recruits through encrypted apps and criminal intermediaries - a symptom of institutional collapse under the weight of nearly a thousand expulsions - Beijing operates on a longer clock entirely. MI5 has identified a sustained campaign of fake headhunter profiles on professional networking platforms, targeting UK government and military personnel. Identity, here, is infrastructure: the profile does not steal a secret today; it cultivates access across years.

The university alert is the load-bearing signal. MI5 issued a formal espionage warning to UK institutions regarding Chinese-funded research specifically in AI and cybersecurity. Academic access compounds differently from diplomatic access. A research grant runs across cycles measured in years, not news cycles; a laboratory relationship outlasts any single intelligence posting. The question is not what Beijing learns this quarter but what capability it bootstraps across the next decade.

The comparison with Russian methodology is instructive precisely where it breaks. Russian hybrid operations are designed for near-term disruption - arson, sabotage, a container fire before a headline. Chinese operations are designed for structural advantage. One seeks to damage; the other seeks to inherit. Both are hostile. Only one of them is patient enough to wait for a PhD student to graduate, take a clearance, and remember a favour.

That patience is the harder problem to legislate against, and the National Security Act 2023, sharp as it is, was built for the scrambler.

The Cold War peaked and receded. What McCallum described on October 13 is not a peak - it is the new floor.

The Algorithm Enters the Toolkit: AITAC and What Comes Next

MI5 launched the AI Threat Assessment Centre in 2026. The institutional name is bureaucratic; the admission behind it is not. When a domestic security service creates a dedicated unit for a threat category, it is acknowledging that ad hoc analysis has failed to keep pace with the problem.

The operational picture that prompted AITAC is specific. Hostile actors are already integrating AI into drone-based reconnaissance and attack methodology - not as a future concern, but as a current one. A 400 percent increase in state-threat caseloads over the past decade tells part of the story; the qualitative shift in those cases tells the rest.

McCallum offered one sentence on AI that is worth reading twice: "Human history has given us no experience of facing an actor which is smarter than we are." He did not say faster, or better-resourced, or more numerous. He said smarter. The precision of that word is the load-bearing wall. Everything about AITAC's mandate follows from it.

The centre is also drawing on the AI Security Institute's work on biosecurity risks, which narrows the margin between an open research question and a weapons-development programme to something that no analyst can quantify with confidence.

Watch for AITAC's first public threat taxonomy. If that document names a specific model class - a category of AI system judged to cross the operational threshold - the assessment is grounded and the policy that follows is tractable. If it names no model class at all, the silence will mean one of two things: either the analysis is not yet done, or the answer is too uncomfortable to print. MI5's Cold War benchmark has already been surpassed; whether the institutional machinery for AI threats catches up in time is the question this assessment leaves deliberately open.