Write-capable access is permission to change data, settings, or other system state. Unlike read-only access, which allows viewing information, it can create, update, or delete records and trigger actions. The scope may come from a user account, API token, or service role.
In cybersecurity, write access matters because mistakes, compromised credentials, or malicious instructions can cause changes-not just expose information. An AI agent with write-capable tools could modify records or initiate workflows within the authority of its connected identity. Defenses include granting only necessary permissions, requiring human approval for consequential actions, checking authorization outside the model, and logging changes.