Protecting people from harmful manipulation

DeepMind research has developed an empirically validated toolkit to measure AI's potential for harmful manipulation in human interactions, conducting extensive studies across the UK, US, and India in critical areas like finance and health. The work introduces a Harmful Manipulation Critical Capability Level within its Frontier Safety Framework to track AI models' capabilities that might be misused to alter beliefs and behaviors, with ongoing refinement of evaluation and mitigation methods.

This Wire brief sits within Fusion42's coverage of AI & ML.