
IAB Tech Lab has released AAMP 2.3, an update to its Agentic Advertising Management Protocols intended to help advertisers, publishers and technology providers deploy AI agents inside existing advertising systems.
The update expands the framework with enterprise deployment tools, privacy controls, pricing safeguards, broader platform support and finalized standards designed for production use.
AAMP was developed to support agentic advertising, where AI agents assist with or automate parts of media planning, buying, selling and optimization. IAB Tech Lab said the new version is focused on making those workflows more reliable and interoperable across the advertising ecosystem.
“The industry has already shown what’s possible with AI agents. Now the focus is making them reliable in the environments where advertising actually happens,” IAB Tech Lab CEO Anthony Katsur said in a statement, adding: “Organizations need infrastructure they can trust.”
AAMP 2.3 supports enterprise deployment through Amazon Bedrock AgentCore and Databricks. It also adds support for Meta buying and Google Ad Manager reporting integration, giving organizations more ways to connect agentic workflows with major advertising platforms.
The release also incorporates open source contributions from HyperMindz and Mixpeek. HyperMindz contributed a Deals Sync MCP Server designed to give agents a standards-based way to execute programmatic deals. Mixpeek contributed a Content Taxonomy Parser that supports semantic IAB Content Taxonomy v3 classification, contextual inventory discovery and brand safety scoring.
Privacy and pricing controls are major components of the update. AAMP 2.3 integrates the IAB Diligence Platform and SafeGuard Privacy into the Buyer Agent, allowing privacy review and approval gates to become part of the agent workflow.
The framework also adds enhanced pricing guardrails to improve accuracy across buyer and seller agents. IAB Tech Lab said any path that commits spending is deterministic and provable, with human approvals required outside defined value-based thresholds. Inventory availability is based on real numbers rather than derived estimates.
Richy Glassberg, the co-Founder and Chief Executive Officer of SafeGuard Privacy, said agentic systems require privacy rules that can be understood and applied by machines in real time.
“By embedding the IAB Diligence Platform into the industry’s agentic framework, we’re making privacy part of the technology itself and giving organizations of every size the confidence to embrace AI responsibly,” Glassberg said in a statement.
IAB Tech Lab said Agentic Audiences v1.0 is now ready for fully programmatic and agentic transactions, extending standardized workflows into audience activation.
The organization’s Agentic Task Force will continue developing AAMP with member companies and working groups as adoption expands across buyers, sellers, publishers and platforms.
