Active Directory succeeded because it gave organizations a unified record of which users and machines existed, who vouched for them, and where their identities were valid. Its model assumes administrators can approve long-lived devices before they join a domain. Cloud infrastructure breaks that assumption: developers and platforms create short-lived instances, databases, Kubernetes workloads, and CI runners faster than humans can register them. Automatically joining these resources with administrator credentials may populate a directory, but it does not provide meaningful oversight. The result is incomplete visibility and weak assurance about which machine identities are trustworthy.
The argument is that modern machine identity has moved from one broad corporate perimeter to many narrow ones, created by service-specific API keys, tokens, and credentials. These can limit the damage from a single compromise, but leaked secrets may enable attackers to move between systems, especially when credentials are stored in plaintext configuration or environment files. Two approaches can help: federation, in which one platform trusts another to authenticate a workload, and credential brokering, in which a secrets manager authenticates workloads and delivers only the credentials they are authorized to use. Federation offers visibility and scoped access but requires many integrations and can leave services that accept only API keys uncovered. A broker can manage those credentials across services. Both approaches aim to restore visibility, trust, and scope without relying on a single directory - a need that becomes more urgent as organizations give AI agents their own identities and permissions.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.