David Adediji

In 2025, the Guardian and the University of Cambridge deployed CoverDrop, a peer-reviewed, independently audited secure communication system built into the Guardian's app. Its architecture is genuinely clever: end-to-end encryption, plausible deniability, and dead-drop message exchange let a source contact a journalist with no traceable link between them, even against a nation-state adversary. It was built for journalists. But the threat model it solves isn't unique to journalism. A child in residential care trying to reach an advocate without their keyworker knowing. A care worker reporting a safeguarding concern outside the management chain they're reporting on. A domestic abuse survivor contacting a solicitor while still living with their abuser. These aren't edge cases — they're daily realities for some of the UK's most vulnerable people, and the secure tools that exist weren't built with them in mind. This talk examines what CoverDrop's architecture can teach us about designing secure communication systems for vulnerable users, and where the hard questions start: which threat model elements transfer cleanly, which assumptions break down outside journalism, and what responsible development looks like when getting it wrong isn't a bug report but a risk to someone's safety.