Vint Cerf, one of the architects of the internet’s core protocols, is backing a new effort to give AI agents a standard way to identify themselves online. According to TechCrunch, Cerf has joined Innovation Labs as an advisor after leaving Google, lending his name and technical credibility to a project aimed at making autonomous software agents more traceable and accountable as they move beyond closed platforms and onto the open internet.

The immediate focus is a proposed standard called DNSid. Innovation Labs, a subsidiary of Identity Digital, says DNSid would tie an AI agent’s identity to an existing domain name and use cryptographic proofs to record registration over time. The pitch is straightforward: if businesses want AI agents to transact, retrieve information, and interact with other systems across the public web, they will need a common way to establish who an agent represents, what authority it has, and who is responsible when something goes wrong.

That matters because much of the current AI agent market still operates inside proprietary environments. Vendors have demonstrated increasingly capable agents, but most are bounded by platform-specific permissions, internal APIs, and product-defined trust models. Moving those systems into a broader, interoperable internet environment raises harder questions about authentication, auditability, and liability. Cerf’s involvement suggests those questions are becoming serious enough to attract infrastructure-level thinking rather than just product-layer experimentation.

What Innovation Labs is proposing

TechCrunch reports that Innovation Labs is trying to build “open architecture” for AI agents to identify themselves. The company’s DNSid proposal appears designed to use familiar internet plumbing rather than create an entirely separate registry stack. By anchoring an agent to a domain, the approach would connect agent identity to an organization’s existing internet presence and administrative controls.

That design choice is also strategic for Identity Digital. As a DNS registry company, it has an obvious interest in extending domain infrastructure into the next layer of internet activity. If more machine-to-machine interaction happens online, domain-linked identity could become a way to make those interactions legible and governable. In effect, Identity Digital is arguing that the domain system could become part of the accountability layer for AI agents.

According to TechCrunch, interim CEO Allie Kline said Innovation Labs is trialing the standards with several unnamed hyperscalers and identity companies. No partners were identified in the reporting, and no deployment metrics, implementation details, or production timelines were disclosed. That leaves the effort at an early stage from an external buyer or builder perspective, even if the company is already engaged in technical discussions.

Cerf framed the project around basic trust questions. In comments reported by TechCrunch, he said the rise of AI agents creates a need to understand what authorities agents hold, where those authorities come from, how identity is established, and why anyone should trust an agent in a given context. Those are not abstract concerns. An agent that books travel, executes procurement steps, accesses private knowledge bases, or negotiates with another service may need to prove not only that it exists, but also that it is acting within valid permissions.

Why AI agents need identity beyond closed platforms

The problem Innovation Labs is targeting is bigger than naming. In today’s product market, many AI agents are effectively extensions of a single vendor stack. A company might run an agent inside its workplace software, customer support suite, or developer tooling, where identity and authorization are inherited from the host platform. That works well enough when the software stays within one company’s boundaries.

The open internet is different. If an AI agent from one company needs to interact with an AI agent or service from another, the systems need shared expectations. Without that, developers fall back on fragmented workarounds: API keys, bilateral integrations, custom OAuth flows, proprietary attestations, or simple trust-by-brand. Those mechanisms can function, but they do not add up to a broadly interoperable trust layer.

Cerf told TechCrunch that one risk is obvious fragmentation: one company’s agent technology may not interoperate with another’s. He compared the challenge to the early internet, where adoption of TCP/IP depended on functionality and user pressure rather than central decree. That comparison does not mean DNSid is on track to become foundational internet infrastructure. It does signal how Cerf is thinking about the problem: standards win when they solve a real interoperability bottleneck better than alternatives.

There is also a governance dimension. A registered domain is relatively static; an AI agent is active, can be delegated tasks, and may act with varying levels of autonomy. TechCrunch quoted Cerf saying it is not yet clear what commitment an organization is making when it registers an agent. That point is crucial. Enterprise buyers may be comfortable proving that an agent belongs to a company, but much less comfortable implicitly vouching for every action it takes.

The politics of infrastructure and trust

Identity infrastructure is rarely just technical. It is also about who controls the registry, who can inspect records, and whether large platform providers use standards to reinforce their own ecosystems. Innovation Labs is trying to position itself as a neutral layer rather than a new AI gatekeeper.

Kline told TechCrunch that the proposal does not come bundled with broader ambitions to control other parts of the AI stack or own the registration data in a proprietary way. She argued that standards launched by a hyperscaler could face resistance if companies fear handing over sensitive operational identity data to a dominant platform. That is a pointed market observation even without naming specific rivals.

For builders and enterprises, the neutrality claim may matter as much as the technical architecture. A standard for AI agents will be difficult to adopt if it is seen as favoring one cloud provider, one model vendor, or one software ecosystem. At the same time, neutrality alone is not enough. Buyers will want to know how DNSid would work with existing identity systems, certificate infrastructure, logging pipelines, policy engines, and incident response workflows.

The article does not answer those implementation questions yet. There is no indication, based on the available reporting, of a finalized specification, governance body, compliance process, or broad industry consortium. What exists now is a proposal, early trials, and a high-profile advisor arguing that identification and accountability are becoming urgent internet problems.

Evidence, claims, and what remains unproven

The core confirmed facts in this story come from TechCrunch’s reporting: Vint Cerf left Google after 20 years, is now advising Innovation Labs, and the group is proposing DNSid as a way for AI agents to identify themselves through domain-linked registration and cryptographic proofs. TechCrunch also reports that Identity Digital owns Innovation Labs and that the company says it is trialing the standard with unnamed hyperscalers and identity companies.

Several important claims remain vendor-reported or unverified. Innovation Labs’ suggestion that domain-name infrastructure is a practical accountability layer for autonomous agents is plausible, but it has not yet been validated by broad market adoption. The value of DNSid will depend on whether major platforms, identity vendors, enterprises, and developers actually implement it in interoperable ways.

Likewise, the notion that more online interaction will eventually happen between agents than humans is presented in the story as a market direction that businesses are envisioning, not an established fact. Cerf himself was cautious on inevitability. TechCrunch quotes him saying he does not think an “agentic economy” is guaranteed, only that people will keep trying to delegate more tasks to software because convenience is powerful.

That restraint is worth noting. The current market has plenty of talk about AI agents, but less evidence of reliable, wide-scale autonomous operation across the public internet. Security, cost control, rate limits, compliance, and error handling remain practical barriers. Any identity layer will have to prove useful in that messy reality, not just in demos.

What this means for builders and enterprise buyers

For product teams building AI agents, the immediate takeaway is that identity is moving closer to the center of system design. If agents are expected to access services, represent organizations, and collaborate across company boundaries, then simple app-level authentication may not be enough. Teams may need stronger provenance, auditable delegation, and revocation mechanisms built into their workflows.

For enterprise AI leaders, the proposal highlights a gap in current deployment models. Many internal AI agent projects can succeed without open-web interoperability, especially when they are limited to systems of record inside one company. But use cases such as supplier coordination, automated procurement, customer support escalation, and cross-company data exchange become harder when each agent speaks a different trust language.

A workable standard could reduce integration friction, but it could also introduce new operational requirements. Enterprises would need policies for agent registration, lifecycle management, delegated authority, monitoring, and dispute resolution. In other words, a standard like DNSid would not remove governance work; it would formalize it.

For the broader enterprise AI market, the story is another sign that the battleground is shifting from model capability alone to infrastructure for reliable deployment. AI agents need more than reasoning and tool use. They need identities that other systems can verify, logs that security teams can audit, and governance models that legal and compliance teams can understand.

What to watch next

The next important signal is whether DNSid moves beyond a proposal into a published specification with public technical documentation. Without that, outside developers and standards bodies will have little basis for evaluation.

A second signal is partner disclosure. Innovation Labs says it is working with hyperscalers and identity companies, but the credibility of the effort will change significantly if major infrastructure providers publicly commit to trials or integration.

Third, watch for governance. If DNSid is meant to become open internet infrastructure, buyers will want clarity on who sets the rules, how disputes are handled, and whether the system can remain neutral as commercial incentives intensify.

Finally, watch whether major AI agents and enterprise AI platforms begin exposing identity primitives meant for cross-platform use. That would suggest the market is preparing for a world where agent-to-agent interaction extends beyond closed software ecosystems.

Creati.ai perspective

Cerf’s involvement does not guarantee that DNSid will become a standard, but it does elevate a real problem the AI industry has mostly deferred. The market has spent the last two years proving that AI agents can act. The next phase is proving that they can act on the open internet in ways that are attributable, governable, and safe enough for real business workflows.

The deeper implication is that AI agents may force a rebuild of some of the internet’s trust layer. If that happens, the winners will not be determined by branding alone. They will be the groups that can connect identity, authority, and auditability to existing infrastructure with minimal friction. Identity Digital and Innovation Labs are trying to make DNSid part of that answer. For now, the idea is early, the trials are opaque, and the market is unconvinced. But the question they are addressing is likely to become harder to ignore as AI agents leave the sandbox.