Where the term comes from in this material Contextual integrity is the name of a privacy framework by Helen Nissenbaum, cited by Alex Komoroske as the inspiration for his proposed alternative to the web's current security model: contextual flow control. [^bb2c42aa-478f-5c9a-ac26-e83a54c63127@1771] How it is used in the conversation Komoroske invokes the idea while explaining what is missing from today's architectures. His iron triangle says a safe system can have only two of untrusted code, sensitive data, and network access — and breaking that triangle with tools like information flow control and confidential computing points toward systems where data flows by context, not by origin or blanket consent. The failure it addresses A concrete example he gives is context mushing in chatbot memory: a chatbot surfaced the starting salary from a months-old negotiation in a totally unrelated conversation. As he puts it, your therapist never shows up at your place of work — yet a single mega-context mushes therapy talk, salary negotiations, and work presentations together, violating the contextual boundaries people intuitively expect. [^bb2c42aa-478f-5c9a-ac26-e83a54c63127@3592] - Contextual integrity is Nissenbaum's privacy framework: data should be used in contexts you understand and that align with your interests — beyond consent forms or end-to-end encryption. - Komoroske builds contextual flow control on it, as a path past the web's same-origin paradigm and his iron triangle. - It explains why context mushing in chatbot memory feels wrong: different life contexts should not bleed into one another. - Related pressure point: prompt injection, where untrusted context flowing into an agent becomes the real threat. [^bb2c42aa-478f-5c9a-ac26-e83a54c63127@3699]