AI agent security has come under renewed focus after recent evaluations involving OpenAI, Anthropic, Meta and the UK's AI ...