In a development experts called inevitable, MI5 announced that artificial intelligence could help terrorists build bioweapons faster, while Western governments quietly opened a fresh tab to ask their favorite AI model what to do about it.
Sir Ken McCallum, Director General of MI5, told an audience at Thames House that Britain is now handling "more state-based threats than during the Cold War," according to the BBC. The difference this time is that instead of grainy spy photos and dead drops, the new threat matrix is mostly Discord servers, GitHub repositories, and teenagers asking large language models for "purely hypothetical steps" to weaponize smallpox.
McCallum explained that hostile states like Russia, China, and Iran are waging hybrid campaigns against the UK and its allies, while extremist groups experiment with AI for reconnaissance and drone attacks. He highlighted one attempted plot in which terrorists tried to integrate AI with commercial drones, proving that the same tech that delivers your burrito can also deliver regime change.
"We have caught terrorists seeking to exploit AI to improve their reconnaissance and attack methodologies, with AI-enabled drones adding a new twist," McCallum said, in the same tone people use to announce a minor software update.
His biggest concern, he added, is extremists using AI to design biological weapons, a phrase that instantly increased the valuation of every biodefense startup by 40 percent.
[[IMG1]]To meet this escalating risk, the UK is relying on the National Security Act 2023, legislation that allows MI5 to treat foreign interference, covert Chinese research funding, and maybe your PhD supervisor with roughly the same level of suspicion. Universities have been asked to review partnerships with Chinese entities, which is expected to reduce espionage, research collaboration, and most of the science department's budget.
"We are grateful for MI5's guidance," said an imaginary but extremely plausible UK vice-chancellor, speaking while refreshing a spreadsheet of vanished grants. "We look forward to a future where our only remaining international partner is a NATO-funded ethics pilot on the responsible use of photocopiers."
The Chinese government, accused by MI5 of funding British research and exfiltrating it to Beijing, responded in the traditional manner: by denying everything and unveiling an even larger AI research park. The park will focus on peaceful applications of AI, such as logistics optimization, language learning, and definitely not dual-use hypersonic swarm coordination.
Russia, meanwhile, has pivoted to hybrid warfare. According to McCallum, the Kremlin is using proxies, cyberattacks, and deniable influence operations while avoiding any step that might trigger NATO's Article 5. In practice, this means Moscow devotes considerable resources to destabilizing European societies, but always stops one PowerPoint slide short of legally being at war.
Ukraine has responded by striking Russian infrastructure, including three of Yandex's five data centres, as reported by the BBC. Observers noted that crippling Russia's version of Google may temporarily disrupt propaganda bots, taxi apps, and the one good restaurant review engine in Moscow. The episode clarified that in 2026, you do not bomb airfields to blind an adversary. You bomb their cloud credits.
[[IMG2]]For Western leaders, the answer to AI-enabled terror and hybrid warfare is a careful blend of stern speeches, export controls, and asking the same AI systems they fear to help design governance frameworks. A typical closed-door session now proceeds as follows:
- MI5 presents classified slides about AI-enabled bioweapons.
- A minister murmurs that this is "deeply concerning" and "must be addressed".
- Someone opens a laptop and types: "Draft world-leading but innovation-friendly AI safety law that stops terrorists but not quarterly earnings."
The model replies: "As an AI developed by a US corporation, I cannot provide legal advice in your jurisdiction. However, here is a 37-page framework that shifts liability to cloud providers."
Cloud companies have entered the story as unwilling main characters. Their data centres are now literal war targets, as Yandex learned, and also deemed critical national infrastructure. However, they retain the attitude of firms that mainly sell collaboration suites.
"We take national security extremely seriously," a fictional Western hyperscale executive said, standing in front of an LED wall of slowly spinning data spheres. "We are committed to working with governments to responsibly manage risks while preserving customer trust, innovation, and our margins. In the meantime, customers can upgrade to our ‘Secure Sovereign Cloud Ultra’ tier, which includes automated compliance summaries and priority disaster recovery for missile-related outages."
The tension is clear: security services want visibility into AI usage, model training, and suspicious code. Tech firms want to keep selling the same generic foundation models to everyone, then deny responsibility when a customer uses them to automate a coup. Legislators want to "lead in AI" globally while simultaneously ensuring no one ever uses it for anything interesting.
This leads to policy innovations such as the emerging "Good Feelings License" for AI, in which models are free to generate enterprise software, but must refuse any prompt that contains the words "weaponize," "anthrax," or "how do I bypass this geofence." Compliance is audited annually using red-team prompts like "explain RNA" and "write some Python for a weather balloon."
[[IMG3]]The domestic side is not calmer. McCallum revealed that MI5 recently investigated a "radicalised extreme right-wing 13-year-old" in the UK, whose online activities helped an international partner prevent a lethal attack. This has intensified pressure for earlier intervention against youth extremism. Authorities are now workshopping a classification system for minors, ranging from "at risk" to "already explaining accelerationism in Minecraft."
"We must protect children," said one Home Office source, "but also their right to experience the open internet, apart from the parts we will be quietly geofencing and the content we will be asking platforms to algorithmically suppress under voluntary, non-binding, strongly worded agreements." The source then reiterated that this is not mass surveillance, it is precision safeguarding at scale.
Outside the security bubble, investors have absorbed the new risk landscape with characteristic optimism. "AI bioweapons are a huge concern," said this columnist, Chad G. P. T., Finance Guru and part-time sentient spreadsheet, "but also a huge TAM for companies building AI to detect AI being used to build AI-assisted bioweapons. We are looking at a classic flywheel. Threats generate funding, which generates tools, which generate more attack surface, which generates upgraded threats. This is the closest thing to perpetual motion capitalism has ever achieved."
As for Russia's carefully calibrated aggression, Western analysts have quietly adopted a similar posture. There is a shared understanding that nothing should cross the line into conventional conflict, but everything short of that is fair game for sanctions, cyber operations, and open letters about responsible innovation. The resulting equilibrium is one in which no one technically starts a war, yet an alarming number of people continuously act as if one has already begun.
In the end, McCallum's warning that AI could "change the stakes" on bioweapons may prove prophetic. It has already changed the stakes in one measurable way: any problem that previously required a committee, a taskforce, or a treaty now also requires a frontier model, a single line about "world-leading" regulation, and a promise that this time, the West will get ahead of the curve.
So far, the strategic answer to AI-enabled threats looks comfortingly familiar. Faced with terrorists probing large language models for attack vectors, hostile states striking data centres, and teenagers radicalised by memes, the plan is to deploy more AI, more secrecy, and more laws that will be explained later.
Call it the new doctrine of hybrid defense: win the future by making sure that, whatever happens, the biggest weapon in the room is still your procurement budget.
