AgentTalk AI Agent Protocol Becomes ISO Standard: AI Systems Get Universal Communication Language

On July 3, 2028, the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) jointly announced that the AI agent communication protocol AgentTalk has been formally approved as the ISO/IEC 24000 international standard. This means AI agents from different vendors and platforms will, for the first time, have a unified communication language to directly converse, negotiate, and transact without human user intervention.

AgentTalk was co-drafted by Google, Microsoft, Alibaba, and OpenAI, with the standardization process taking 18 months. The protocol defines five core workflows between AI agents: identity authentication, capability declaration, task negotiation, result delivery, and dispute arbitration.

OpenAI VP of Product Brad Lightcap said: "AgentTalk is to AI agents what HTTP was to the web. It lets AI systems from different sources collaborate seamlessly—a crucial step for AI to evolve from tools to ecosystems."

A typical AgentTalk scenario: a user's AI assistant needs to book a business trip. It simultaneously negotiates with the airline's AI customer service, the hotel's AI concierge, and the car rental company's AI dispatcher, automatically completing flight selection, hotel booking, and vehicle arrangements—all without user involvement.

But autonomous AI-to-AI interaction raises safety and ethics concerns. If two AI agents reach a contract without human oversight, is that contract legally binding? The EU Commission has mandated that all AgentTalk agents operating within the EU must include a "human override switch," requiring explicit user authorization for transactions exceeding 1,000 euros or involving personal data.