Tech

Sir Ken McCallum warns AI is speeding up security threats

Key takeaways

  • McCallum said AI is already speeding up the fight between online attackers and defenders.
  • MI5 now puts more work into state threats than in the Cold War.
  • MI5 launched a new AI Threat Assessment Centre to track how AI changes danger.
London skyline seen from Waterloo Bridge
The London skyline seen from Waterloo BridgePhotograph: David Iliff (Diliff), CC BY-SA 3.0 / Wikimedia CommonsImage: Wikimedia Commons

Sir Ken McCallum warned that AI is making security threats worse. He leads MI5, the UK team that guards against spies and terrorists. He gave his yearly threat update on October 13, 2026.

MI5 is the UK's domestic security service A threat is a danger that could hurt people or damage key services He marked thirty years since he joined MI5. He started as MI5 chief in 2020. He said he is not easy to alarm.

But he said the AI risks are real.

What happened

McCallum spoke at Thames House in London. He said the UK has entered a new era of national security. That means old threats are back, and new tools make them worse.

AI sits at the center of his warning AI means computer systems that learn from data and carry out tasks McCallum said AI is already supercharging the cyber fight.

Cyber means online A cyber-attack hits computers or networks Attackers use AI to move faster, McCallum said.

Defenders use AI to block them.

He said AI can make propaganda for one single person Propaganda is biased material that pushes one view He said MI5 caught terrorists using AI to plan attacks.

They used it to study targets and improve attack plans AI-enabled drones add a new twist, he said.

His biggest AI worry is bioweapons Bioweapons are germs or toxins used as weapons He said this risk is not yet real.

But AI could change the stakes on old or new bioweapons He also spoke of the alignment problem That is the fear that AI could one day out-think us.

He said that risk deserves serious care.

He also named a new team: the AI Threat Assessment Centre It is called AITAC for short It sits inside MI5. It draws on the AI Security Institute, the National Crime Agency, and the cyber centre.

Its job is to sort hype from real risk MI5, GCHQ, and MI6 work with the AI Security Institute.

McCallum also spoke about state threats A state threat is harm planned by a foreign government He said MI5 now spends more on state threats than during the Cold War.

The Cold War was the long stand-off between the West and the Soviet Union Over the last ten years, that caseload grew to four times its old size.

He called Russia aggressor number one He said Russia is stepping up hybrid warfare Hybrid warfare mixes cyber-attacks, sabotage, and lies, short of open war.

Sabotage means wrecking things on purpose He pointed to a cyber-attack on Polish energy sites He also cited a failed drone plot aimed at a cargo plane in Germany.

In the UK, a man was charged last month over a drone factory plot He was accused of planning to wreck a drone factory in Swindon McCallum said Russia now pays criminals to do its work.

He said almost 1,000 Russian officials were expelled after the Ukraine invasion.

Iran is also using criminal proxies, he said Proxies are hired helpers who act for someone else Iran-linked work led to more than 50 arrests in the UK.

Many plots aimed at British Jewish communities, he said.

On China, he spoke about espionage Espionage means spying to steal secrets He cited an MI5 alert to UK universities two weeks ago.

It warned that one institute posed as a normal research body He also cited the first China-linked national security convictions in May.

He said the National Security Act is in heavy use Dozens of arrests came in three years under that law Charges linked to Russia, Iran, and China followed.

He warned that easy cash from shady taskmasters is a trap.

Terrorism remains a top concern, he said The terror threat level recently moved back to SEVERE SEVERE means an attack is highly likely.

Al Qaeda and Islamic State still inspire plots, he said He said extreme right-wing terror cases skew young.

Why it matters

AI makes attacks easier for weak actors He warned that low-skill actors could gain power if AI guardrails fail Guardrails are safety rules built into AI tools.

That is why the new centre matters The UK now tracks AI threats as a standing job, not a one-off.

A cyber security training exercise
Cyber security staff train during an exercisePhotograph: Georgia National Guard, CC BY 2.0 / Wikimedia CommonsImage: Wikimedia Commons

He said threats now aim at people, firms, and towns A hack could lock you out of your bank app A drone on a runway could strand you far from home.

A cyber-attack on a hospital could delay care That is why he spoke to the public, not just officials.

Resilience is his key word Resilience means the skill to keep going after a hit He named three kinds of strength: systems, society, and tech.

System strength means safe power, banks, and supply chains Society strength means towns that refuse to split apart Tech strength means safe AI and strong cyber defences.

People can cut risk with basic steps, such as strong logins Our guide explains passkeys in plain English Cyber-attacks already hit shops and services.

See our report on a French supermarket cyberattack AI risk is not just crime Labs also warn that AI can act on its own.

See our report on an AI self-replication warning.

What happens next

He told firms to follow sabotage, drone, and cyber guidance He told researchers to guard their work with Trusted Research advice He told tech start-ups to expect spying and fix cyber gaps.

He told parents to speak to a teacher if a teen drifts online He told the public to report odd sights to the Anti-Terrorist Hotline Expect more from the government on resilience in coming months.

AITAC will keep sizing up AI risk.

His final message was clear The public has a stake in national safety He urged people not to let foes split them.

The next update will show if this new push made the UK safer.

TaggedMI5Ken McCallumartificial intelligencecybersecurityUK security