MI5 warns AI could widen terrorists’ access to bioweapons
The warning came as ministers outlined new measures to test advanced AI and prepare for serious incidents.
Follow-up to: Ministers expand AI defences against cyber attacks Monday, 12 October 2026, 20:41
Update
MI5 director-general Sir Ken McCallum said Iran was seeking to target military sites in the UK and that the suspected plot at RAF Fairford was not unique. Seven men in their 20s had been arrested and released on bail under investigation. (Associated Press)
MI5 director-general Sir Ken McCallum has warned that artificial intelligence could increase terrorists’ and hostile states’ ability to cause harm, according to a report published on Monday. The warning adds a national-security dimension to the government’s new programme for testing advanced AI systems and strengthening cyber defences.
Brit Brief reported that McCallum told ministers artificial intelligence was among the most important issues for Britain’s future security. The report said intelligence chiefs were concerned about hostile states, including Russia and Iran, as well as terrorist groups using AI to improve their capabilities.
The warning is about potential misuse rather than a confirmed attack. The available reporting does not identify a live plot, name a group that has obtained a biological weapon through AI, or establish that an AI system has independently produced one. The distinction matters because the underlying concern is about access to knowledge and assistance, not evidence of a completed operation.
The assessment follows a government announcement on 12 October that Britain would expand pilot projects using frontier AI to find vulnerabilities in public-sector systems. Ministers also said they would consider stronger reporting requirements for serious AI incidents and near misses, while preparing for incidents involving autonomous AI agents.
The government’s announcement said the AI Security Institute was evaluating the latest generation of advanced models. It described possible risks from systems that act with limited human supervision, but did not claim that such systems had already escaped human control.
Anthropic’s September threat-intelligence report provides separate context for the biological-risk discussion. The company described five cases in which its models were used for work that could support biological research with harmful applications. It withheld the identities of the people and institutions involved and stressed that model evaluations cannot demonstrate how a capability would be used in the real world.
Britain’s terrorism threat level remains severe, meaning an attack is considered highly likely, although that assessment covers terrorism generally and is not an AI-specific warning. MI5 has not publicly said that the threat level was changed because of the latest comments.
The policy question is how much responsibility should sit with technology companies, government agencies and users. The warning supports a case for stronger testing, reporting and safeguards, but the accessible sources did not include a full MI5 transcript or independent evidence of a specific biological-weapons capability. The precise basis for McCallum’s reported comments therefore remains incomplete.
One story, several perspectives
What is established
- MI5 has warned that AI may increase risks from terrorists and hostile states.
- The government announced new AI testing, cyber-defence and incident-preparation measures on 12 October.
- Anthropic has reported potentially harmful biological-research uses of its models while acknowledging uncertainty about real-world effects.
Left
Arguments Advanced AI should face strong public safeguards, mandatory incident reporting and independent testing before release. Companies should not decide alone what level of biological or security risk is acceptable.
Values Collective security, precaution, public accountability and protection from private power.
Consequences Supporters fear that weak regulation could leave the public exposed, while stronger controls could reduce misuse and make companies responsible for foreseeable harms.
Centre
Arguments Britain needs proportionate rules based on evidence, combining government oversight with technical expertise and cooperation from companies. Regulation should target serious risks without blocking beneficial research or innovation.
Values Institutional competence, transparency, proportionality and a balance between safety and economic opportunity.
Consequences This approach could produce workable safeguards, but may be criticised if regulators move too slowly or if responsibilities remain unclear during a crisis.
Right
Arguments Security agencies should focus on concrete terrorist and state threats rather than broad restrictions on AI development. Targeted controls, intelligence work and strong penalties for misuse may be preferable to heavy regulation of an entire industry.
Values National security, innovation, competitiveness and limits on state intervention.
Consequences Supporters fear that sweeping rules could weaken Britain’s technology sector, while critics would worry that voluntary safeguards leave dangerous gaps.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved with corrections · Priya Shah — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The government measures, Anthropic report and current threat level are supported by primary sources. The specific MI5 warning is supported by one accessible independent report, but its full primary wording was unavailable.
- confirmed Sir Ken McCallum warned ministers that AI posed a major future national-security concern. — Reported by Brit Brief; the supplied headlines show the warning was also covered by several UK outlets. source
- confirmed The warning concerned hostile states and terrorist groups using AI. — Brit Brief attributes this assessment to intelligence chiefs, including McCallum. source
- confirmed The government announced expanded AI cyber-defence work on 12 October 2026. — GOV.UK announcement dated 12 October 2026. source
- confirmed Anthropic reported five cases involving potentially harmful biological research uses of its models. — Anthropic’s September 2026 threat-intelligence report describes five case studies. source
- confirmed The UK terrorism threat level is severe. — MI5’s current threat-level page defines severe as an attack being highly likely. source
- uncertain AI has already given terrorists access to a biological weapon. — No source used establishes this; the article explicitly removes that implication. source
1 correction(s) applied
- Was: AI could give terrorists access to bioweapons.Now: AI could increase terrorists’ access to knowledge and assistance relevant to biological weapons. (The available sources support a warning about potential misuse, not a confirmed capability or completed attack.)
Editor's note
The warning is reported by multiple outlets in the supplied feed and by an accessible independent report. The full primary MI5 statement was not available by the deadline, so no specific plot or confirmed bioweapons capability is claimed.Sources
The story so far
- Monday, 12 October 2026, 20:41 Ministers expand AI defences against cyber attacks
- Tuesday, 13 October 2026, 14:35 MI5 warns AI could widen terrorists’ access to bioweapons (this article)
More on this in other media
- The Guardian — „artificial intelligence”
- BBC News — „artificial intelligence”
- Sky News — „artificial intelligence”