Your CI pipeline is about to start trusting machines you cannot stare in the face. The agent that opens a pull request, the one that runs the deploy, the one that signs the artifact: each needs a name your build steps can verify, and most teams still resolve "which agent" by reading a bot account's handle. That is not identity. That is hope with a profile picture.
This week the Linux Foundation declared its intent to launch the Agent Name Service, an open standard for giving AI agents verifiable identities by tying them to DNS. The Foundation made the call on Tuesday, per reporting in The New Stack, and the bet is unsubtle: the namespace that already proves "this is example.com" is the namespace that should also prove "this agent works for example.com."
Where the spec came from
ANS did not start at the Foundation. The first version was a research paper published in May 2025 by the OWASP GenAI Security Project, co-authored by Ken Huang of DistributedApps.ai and Akram Sheriff, an AI security engineer at Cisco. A second version landed in April as an individual draft at the Internet Engineering Task Force, and that revision is where the design pins each agent to a real domain its operator already controls.
