Claude's API can pin safety actions to one end customer
User profiles tell Anthropic which end customer a request is for, so a restriction can hit that profile instead of the whole organization. Enterprise inference hooks are also out of beta.
Official source (opens in a new tab)
Anthropic's Claude Platform release notes for 20 October 2026 launch user profiles. A platform, gateway, or reseller can tell Anthropic which end customer each API request belongs to. Anthropic says a safety enforcement action can then apply to that one end customer instead of the whole organization. The SDK and the ant CLI support the new /v1/user_profiles endpoints.
The same notes take inference hooks out of beta for Claude Enterprise organizations. Organization webhooks are new on the Admin API, in beta, and need the webhooks-2026-10-07 header. The /v1/organizations/webhooks endpoints cover two events: Anthropic restricts or bans one of the organization's user profiles, and that restriction or ban ends. Anthropic says to ask your account team for access.
Who it's for: a company that resells or proxies the Claude API and does not want one customer's enforcement action to shut off everyone else.
Fits: Best for Writing stays Claude for the draft. These are API administration controls. The writing verdict is unchanged.