Models and misuse
AI capabilities
Examine new model capabilities and how they could be used in attacks or defence.
NWQ Cybersecurity
We study how AI changes attacks, defences and the systems organisations depend on.
Research
Models and misuse
Examine new model capabilities and how they could be used in attacks or defence.
Agents, data and access
Examine where AI systems can expose data, misuse permissions or act beyond their intended scope.
Controls and trade-offs
Compare possible responses, their costs and the risks that remain.
Work with us
Each piece of work starts with a specific system, decision or research question.
A briefing on AI developments relevant to your organisation, with sources and questions to consider.
An assessment of how AI could change attacks against a particular system or sector.
Work through a possible incident to test assumptions, responsibilities and response plans.
An investigation of a specific AI security question.
Method
Read the research and examine reported attacks.
Describe how an attack could work and where assumptions are involved.
Check claims against evidence and consider possible responses.
Set out findings, limits and options.
Contact
Tell us about the system, decision or research question you want to examine.
Checking the contact form… You can also email us directly.