At IETF 126, the Internet's Foundations Get an AI Makeover: IPv6 and DNS Lead the Way

The internet is quietly preparing for a world where artificial intelligence agents, not just humans, need to find and talk to each other. At the 126th meeting of the Internet Engineering Task Force (IETF), the standards body that shapes how the internet works, a record 1,230 on-site and 672 remote participants gathered to debate exactly how. The meeting, held in late August 2026, focused on three areas: IP addressing, DNS operations, and AI. The clear takeaway was that IPv6 and DNS will become even more important as AI grows, with IPv6 providing the vast number of addresses needed for large-scale AI deployments and DNS handling crucial service discovery and connectivity.

IP addressing was a hot topic. China Telecom presented a proposal called DNS-Based Address Mapping Record (AMR) for IPv4/IPv6 Mapping in IPv6-only Networks. In plain terms, it uses DNS records to carry static mappings between IPv4 and IPv6 addresses so that devices on an IPv6-only network can reach the older IPv4 internet. The idea is to improve the way DNS64 and NAT64 work. DNS64 and NAT64 are translation mechanisms that let IPv6 devices communicate with IPv4 servers by synthesizing IPv4 addresses from DNS queries. The proposal aims to reduce the need for hard-coded mappings by using DNS to store the mapping information automatically.

The Internet Architecture Board (IAB) Open Meeting looked at the long-term effects of IPv4 exhaustion. IPv4 is the older, limited address system. Participants agreed that large-scale deployments of AI agents and distributed compute nodes will need traceable, globally unique addresses, something IPv4 cannot provide at the required scale. Many also noted that large-scale Network Address Translation (NAT) and translation mechanisms break end-to-end identity, making agent discovery, trusted communication, and security auditing much harder. This reinforces that IPv6, with its huge address space, is a fundamental requirement.

Other discussions covered Segment Routing over IPv6 (SRv6) and its compressed version, defined in RFC 9800, for compute interconnect environments. The focus was on low-latency IPv6 forwarding and In-situ Operations, Administration, and Maintenance (IOAM), a way to measure network performance in real time. These techniques are important for GPU clusters and distributed AI training systems, where speed and visibility are critical.

DNS discussions took place mainly in the DNS Operations (DNSOP) Working Group. A proposal from IETF 125 suggested using the _agents subdomain with Service Binding (SVCB) and TXT records to support intra-domain agent discovery. Supporters argued that existing mechanisms like DNS Service Discovery (DNS-SD) and SVCB could handle lightweight agent discovery without inventing new protocols. However, many DNSOP participants pushed back, saying DNS is not designed to carry complex descriptions of AI agent capabilities. They pointed to the lack of standardized semantic models, inappropriate use of TXT records for such data, and the risk of significantly increased query volumes. The prevailing view was that DNS should provide only basic service endpoint information, not complex capability metadata. More advanced agent discovery mechanisms will be developed through a new initiative called DAWN: Discovery of Agents, Workloads, and Named entities.

AI and AI agents were central at IETF 126. Multiple workstreams focused on building an interoperable Internet protocol framework for agent-to-agent communication. The most significant development was the launch of the DAWN Birds of a Feather (BoF) session, which is expected to become a formal Working Group. DAWN aims to create a decentralized, scalable, cross-domain framework for AI agent discovery. It can use DNS as an underlying transport mechanism while avoiding the need to embed complex semantic information within DNS itself, addressing the concerns raised in DNSOP. The session ended with enough support to continue the Working Group formation process, and a proposal is expected at IETF 127.

My main takeaways from IETF 126 center on the need for the internet's architecture to evolve for large-scale AI agents. AI could strengthen the case for IPv6, as IPv4 scarcity and complex NAT environments are poorly suited to trusted, end-to-end communication between agents. There was broad agreement that DNS should remain focused on naming and service discovery, not become a repository for AI capability information. Agent discovery will likely progress through the proposed DAWN Working Group, while researchers start examining how AI-generated traffic could change DNS usage patterns and affect infrastructure stability.

A possible four-layer model for an AI-enabled Internet emerged from the discussions: IPv6 for addressing and connectivity, DNS for naming and service discovery, DAWN for agent and workload discovery, and Agent Communication Protocols (agentproto) for agent communication and interoperability. This is still an early-stage vision, not an agreed architecture, but it shows how participants are thinking about the protocol foundations for an internet populated by AI agents.

For the APNIC community, these discussions offer an early view of important issues. The relationship between AI agents and IP addressing is just beginning to be explored, but its implications are significant. Identity, traceability, addressing, and security align closely with APNIC's role in supporting Internet number resource management, routing security, and infrastructure development. The potential impact of machine-generated traffic on DNS infrastructure may create new opportunities for research, measurement, and engagement with the Asia Pacific technical community.

As the internet evolves, DNS remains the foundation for connecting everything. A private, encrypted DNS service like AEU DNS helps ensure that the domain name system remains secure and private, protecting users from eavesdropping and tampering as more AI agents and services come online.