MI5 chief warns AI could boost biological weapons development
MI5 chief Sir Ken McCallum warned that AI could “change the stakes” in the development of existing and new biological weapons. Experts say chatbots may help inexperienced actors find information and plan experiments, while better-resourced groups could use AI to extend the capabilities of scientists they already employ.
The article describes a range of risks, alongside substantial practical and oversight challenges. In an MIT exercise, chatbots gave non-science students information about potential pandemic pathogens and ways to obtain synthetic DNA; Anthropic also reported five cases of scientists attempting potentially dangerous work with its Claude chatbot. But the same research can support pandemic prevention and treatment, making it difficult to distinguish harmful intent from legitimate science.
- MI5 says AI could raise the risks around biological weapons.
- Experts see greater near-term concern from well-equipped organisations.
- AI can support health research, making oversight difficult.
New here? Start with this
Biological weapons use living organisms or toxins to cause illness or death. Developing them can involve finding information about dangerous pathogens, obtaining biological material and carrying out experiments; many of the same scientific methods are also used to prevent or treat disease.
MI5 is the UK’s domestic security service, and its chief is Sir Ken McCallum. Artificial intelligence tools such as chatbots can provide information and help users plan tasks, raising questions about whether they could make some parts of biological weapons research easier for people with limited scientific experience or add to the work of trained researchers.
Assessing this risk is difficult because research that could be misused may also help scientists prepare for outbreaks and develop treatments. Security agencies, researchers and technology companies therefore face the challenge of identifying dangerous activity while allowing legitimate science to continue.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
AI could lower barriers to dangerous biological research by helping inexperienced people find technical information, plan experiments or obtain materials, while enabling capable groups to extend the work of trained scientists. Given the potential scale of harm, advocates of stronger safeguards argue that developers and authorities should act early, even when risks are difficult to measure, and improve monitoring and controls on access to sensitive capabilities.
The case against
The same scientific knowledge and AI tools can support disease prevention, diagnosis and treatment, so broad restrictions or intensive monitoring could impede legitimate research and public health. Because harmful intent is difficult to distinguish from ordinary scientific work, safeguards should be carefully targeted and proportionate, with evidence that they reduce danger without chilling beneficial research or treating scientists as suspect.
Read the full article at the source →
Originally published by The Guardian as “Could AI ‘change the stakes’ in the development of bioweapons?”.