AI Trust: New Standards For The AI Era

AI Can Infer Meaning. It Cannot Create Trust.

The next internet will not be built by humans clicking buttons.

It will be built by AI agents acting on behalf of people, brands, publishers, platforms, and enterprises. They will search, negotiate, buy, optimize, measure, and decide. Some will be simple workflow assistants. Others will become increasingly autonomous economic actors managing budgets, audiences, inventory, and outcomes across systems that were never designed to work together.

This raises a question that is becoming increasingly common in AI circles: If large language models can understand context, infer meaning, and reason across messy data, do we still need standards and protocols?

It is an appealing argument. After all, much of the internet’s infrastructure exists because software historically struggled with ambiguity. We built schemas because machines could not interpret intent. We built taxonomies because systems could not reconcile different definitions. We built APIs because applications could not understand one another without rigid interfaces.

Today, an LLM can often bridge those gaps automatically. It can recognize that “campaign start date,” “flight begin,” and “launch timestamp” all describe the same concept. It can translate schemas, map taxonomies, generate integration code, and reason across incomplete information.

Perhaps standards become less important when machines become smarter. The argument is directionally right. And dangerously incomplete. The mistake is assuming that standards exist primarily to help machines understand information. They do not. Standards exist to help markets coordinate activity.

Trust: The Most Important Infrastructure Challenge Of The AI Era

For decades, the advertising and technology industries invested heavily in interoperability. We standardized bid requests, consent strings, audience taxonomies, measurement events, reporting formats, and identifiers because software required structure.

LLMs fundamentally change that equation. They are exceptionally good at interpretation. They can act as semantic middleware between systems that were never designed to communicate. They can infer intent from incomplete metadata and resolve ambiguity that previously required human intervention.

But meaning and trust are not the same thing:

• A model can infer what a field probably means.
• It cannot prove whether the data was authorized.
• It cannot prove whether a signal originated from a legitimate source.
• It cannot prove whether an action remained within delegated authority.
• It cannot prove whether accountability exists for the outcome.

Those are not language problems. They are trust problems. And trust is rapidly becoming the most important infrastructure challenge of the AI era.

AI Trust: The Old Internet Verified Access. The Next Internet Must Verify Agency.

Most existing control systems answer relatively narrow questions:

• Identity asks who logged in.
• Consent asks whether data use was permitted.
• Fraud systems ask whether activity appears invalid.
• Measurement asks whether an event occurred.

The agentic internet introduces a more difficult question: Did this action remain bound to an accountable principal throughout its lifecycle?

That principal may be a person, an organization, a role, or a delegated authority structure. The critical issue is not simply whether an action happened, but whether the action remained connected to the authority that permitted it.

This distinction becomes increasingly important as AI systems begin operating independently. AI can already generate valid-looking behavior at scale. It can create plausible audiences, customer journeys, optimization decisions, campaign recommendations, and transactional activity. Much of this behavior may be useful. Some may be fraudulent. Much will occupy a gray area between automation and accountability.

Reasoning Helps Systems Talk. Trust Helps Markets Function.

This distinction is easy to miss. LLM reasoning is exceptional at interpretation. Protocols are essential for coordination:

• A model can resolve ambiguity. Markets cannot settle transactions on ambiguity.
• A model can explain why it reached a conclusion. A protocol determines whether it was authorized to act.
• A model can infer that two audience definitions are similar. It cannot independently determine whether either audience was created lawfully, remains valid, or should be trusted.

The issue was never machine comprehension alone. The issue was coordination cost, accountability, and trust. That is why standards continue to matter. Not because machines cannot understand information. Because markets cannot function without shared rules.

The Standard Moves Up The Stack.

The future will not look like the past. We will likely need fewer taxonomy wars, fewer rigid schemas, and fewer brittle integrations. AI can absorb much of that complexity.

What remains are the questions inference cannot solve:

• Questions of provenance.
• Questions of permission.
• Questions of delegation.
• Questions of authority.
• Questions of continuity.
• Questions of accountability.

AI Trust Becomes Runtime Infrastructure.

Historically, trust has been retrospective. A campaign runs. A transaction settles. An audit follows. Trust is evaluated after the fact.

Agentic AI systems change that model. As AI agents begin negotiating deals, configuring campaigns, selecting audiences, allocating budgets, and making optimization decisions, trust can no longer be evaluated solely through audits and reporting. It must increasingly be evaluated at the exact moment an action occurs.

That may sound subtle, but it represents a fundamental shift in market infrastructure. Trust moves from an audit function to an execution function. From governance review to runtime control. From post-campaign analysis to real-time decisioning.

This is where much of today’s discussion around protocols misses the bigger picture. The future is not simply about how agents communicate. It is about how agents earn the right to act.

Advertising Will Feel This First.

Advertising is one of the first industries where agentic trust becomes commercially important.

Media markets already trade on signals that claim to represent human attention, intent, audience quality, and commercial context. AI makes it increasingly easy to generate signals that appear valid while remaining disconnected from accountable origins:

• A bid request may look legitimate.
• An audience segment may appear valuable.
• An optimization may improve performance.
• A campaign may deliver results.

Yet none of those outcomes answer a more important question: who authorized the action, under what authority, and within what scope? That is not simply a fraud problem. It is a market-design problem. Programmatic advertising has always depended on trust in signals. The next phase will depend on trust in actions.

The future is not a choice between protocols and AI. Protocols without AI become bureaucracy. AI without protocols becomes persuasive chaos. The winners will combine both: systems where models handle interpretation and adaptation while standards govern authority, accountability, and trust.

The first internet standardized information. The second standardized transactions. The third will standardize trust. Not because AI is weak. Because AI is becoming powerful enough that trust must be verified before action, not after it.

Don’t miss any important news from the world of digital marketing, both nationally and internationally! Subscribe to the new DMEXCO newsletter “Digital Digest” and receive the latest information directly in your mailbox every Monday, Wednesday, and Friday.