How company data is kept apart
Several independent checks stand between one company's data and another's.
- The server decides which company you are working in from your membership. The browser cannot choose it.
- Every database query for company data names the exact company and workspace it may touch.
- The application connects to the database with its own restricted login. It owns nothing, so it cannot change the database structure, switch off protections or delete audit history.
- For many record types, the database itself refuses a link from one company's record to another company's record.
- One central permission layer decides what each person may read: their sections, the company's module switches, who a record is shared with, and field rules such as who may see pay.
- An automated leak test runs two test companies through every role against moss's read paths, exports and search. Any data from the other company fails the test. It is part of the release check we run before publishing a change.