Identify the behavioural patterns causing AI risk

The Signals Library helps you recognise when behaviour is introducing risk into AI  workflows. Each signal outlines what to look for, the research questions to ask, and the of data needed to measure it. Organised into 14 behavioural signals, each one helps your team distinguish between normal usage and emerging risk. 

Understand the consequences of high risk behaviour.

The Impact Assessment helps teams identify where actual behaviour may lead to poor outcomes. It connects how AI is used to specific failure modes, evaluating the impact on performance, compliance and safety. This tool helps you design targeted responses to reduce behavioural risk before it scales. 

Define how AI should be used.

This tool brings together product design with risk and governance to articulate what safe behaviour looks like, and the thresholds beyond which behaviour becomes risky. The tool helps teams explore key risk scenarios to create a shared understanding of how to recognise risy behaviour and which controls to use to nudge behaviour back into safety. 

Validate behaviour in the real world.

The experiment lab provides a structured way to design and run experiments that test how AI is actually used. Teams capture key data, observe real life behaviour, and interpret their findings to identify and manage behavioural risk. By linking real-world actions to known risk signals, you get a clear view of what works in practice, not just in theory.