LUCAMETwo separate issues have surfaced this week around AI language models. One involves cached data in the GLM-5.3 model being accessible in ways it shouldn't be, and the other is a technique called context bombing, where attackers overload an AI's context window to manipulate or extract information from it.
What this means for your business
If your business uses AI assistants, chatbots or any AI-powered tools that handle client data or internal documents, these aren't theoretical risks. You should check with your IT provider whether the AI products you use store or cache prompts and responses, and whether those are isolated per user or session. The ICO expects businesses to treat AI-processed personal data with the same care as any other system, so a gap here could mean a reportable breach.