New NAI Guidance: Key Do’s & Don’ts for Using AI in Network Advertising
A set of plain-English do’s and don’ts for adopting AI-enabled and agentic advertising workflows.
Today, the NAI’s released new guidance, Key Do’s & Don’ts for Using AI in Network Advertising, that is designed to help ad-tech companies keep their privacy and data governance practices up-to-date as AI systems used in advertising gain more capabilities and authority to act. It gives members a set of do’s, don’ts, and working questions they can apply to their own AI systems today.
AI is not new to digital advertising. NAI members have used machine learning, optimization, and predictive modeling for well over a decade, and the NAI’s Self-Regulatory Framework provides baseline privacy expectations for AI use in digital advertising. What is new are the generative capabilities of some AI systems and how much they can do on their own.
AI tools can now turn a plain-English prompt from an advertising campaign planner into an audience segment. Agentic systems are beginning to access data, bid on inventory, and execute transactions with more autonomy. NAI members are deciding how to use these capabilities while technical standards and market practices are still taking shape. These developments raise an important practical question: how should privacy and data governance keep up as AI systems gain more capabilities and authority to act? The NAI’s new guidance provides critical answers to this question.
What’s in the guidance
This guidance covers nine topics, each with practical do’s and don’ts:
- Inventory and enabling of AI use cases
- Advertising audience/segment review and activation
- Testing and monitoring of AI systems
- Disclosures about how AI systems are used
- Permissions and constraints applied to AI systems
- Choice and signal handling
- Oversight and logging for agentic AI systems
- Contracting and risk allocation between AI users and AI vendors
- Accountability
The guidance also includes a one-page checklist that turns the topics covered into working questions. Do we know which of our tools are AI-enabled, and who owns each one? Would we notice if an AI-built segment turned out to be a proxy for a health condition? If a consumer opts out, does that choice still travel with the data when an agent acts on it?
Proportionality
The recommendations made in this guidance should be interpreted to scale with a system’s complexity, capabilities, and autonomy. An advisory tool that surfaces options for human review calls for proportionally lighter treatment than an AI agent that can act before anyone understands the specific decision. The more authority a system has to change a privacy or legal outcome on its own, the stronger the case for testing, permissions, monitoring, and a proven way for human intervention. In every case, people remain accountable for the results.
What’s next?
This voluntary guidance is now available to use. Members and other stakeholders can use it to review current workflows, spot gaps in controls or ownership, and focus oversight where it matters most. The NAI will continue monitoring the incorporation of AI into digital advertising as technology and standards evolve, and look for opportunities to provide further guidance.
For more information, contact media@thenai.org.