A year of AI coding-agent incidents.

Coding agents now read your source, hold your credentials, and can reach production. In the last year, plenty has gone wrong in public: private code and secrets shipped to outside servers, production databases wiped in seconds, supply-chain worms that turn installed agents into the thing hunting for keys. The fallout is rarely just technical. It is hours of downtime, lost customer data, leaked intellectual property, breached trust, and real damage to a company's reputation. We pulled the most significant incidents into one place, each linked to its source.