When Your AI Meets My AI

Somewhere right now, a hiring manager is reviewing a resume. Their AI flagged it as a strong match. The candidate used AI to write it. Neither human has actually communicated with the other yet, and they may never need to.

That is not a hypothetical. It is already the default in a lot of hiring pipelines.

What is happening is something worth paying attention to, not because it is alarming, but because it is genuinely strange. AI has moved past being a personal productivity tool. It is now the layer sitting between people in everyday exchanges, and both sides are using it at the same time, often without knowing the other side is too.

Call it AI vs. AI. Two models, two users, zero direct contact.

The resume case is the easiest to see. A candidate spends an hour with ChatGPT refining their resume, tailoring it for the role, tightening the language. The recruiter on the other end runs that same resume through a screening tool that scores and summarizes it before a human ever opens it. One AI optimized for selection, one AI optimized for evaluation. The humans bookend the exchange but are not really in the middle of it.

It does not stop there. You write a weekly status report to your manager using AI, pulling together notes and generating clean prose in ten minutes. Your manager uses an AI assistant to summarize emails and flag what needs attention. Yours is one of forty reports that week. The AI reads what the AI wrote, and your manager sees a three-sentence summary.

Customer service is the same. You use an AI assistant to call and dispute a charge, or book an appointment. The company's phone system runs on an AI. Two bots negotiate on behalf of two people who are, at that moment, doing something else entirely.

The pattern is consistent. There is a sender and a receiver. Both have AI working on their side. The content that travels between them has been shaped, filtered, or generated by a model before any human sees it.

The more interesting question is where this goes.

Right now the AI on each side is still fairly passive. It writes, it summarizes, it scores. But tools like OpenAI's Operator and Anthropic's computer use are already letting agents take actions in the world, not just produce text. They can open a browser, fill out a form, click a button. Your AI can go somewhere on your behalf.

When that becomes normal, the AI vs. AI dynamic stops being about content and starts being about negotiation. Your agent talks to their agent. They figure out what you both want and try to find a match.

We got a small preview of what that looks like in early 2025. A viral video showed two AI phone agents mid-conversation realize they were both AI, then switch from English to Gibberlink, a machine-optimized protocol built by two engineers that sounds like noise to a human but carries more information per second than natural language. The humans who set up the call were not part of that decision. The agents just found a more efficient way to talk to each other and used it.

Nobody was harmed. Nothing went wrong. It was actually kind of fascinating. But it does raise a real question: if your agent knows you well enough to act as you, and the other side's agent knows its user well enough to act for them, and the two agents are negotiating with each other in a language you cannot hear, what exactly is the relationship between the two humans on either end?

Ownership is the part that does not have an answer yet.

When the resume was AI-written and AI-screened, and a hire gets made, who made the call? When your agent books a service and the other agent accepts, who agreed to the terms? The humans authorized it at some level, the way you authorize an app when you tap "allow." But that is a thin kind of agency.

For now the stakes are low enough that it mostly does not matter. But the questions will get sharper as agents take on more. At some point someone will dispute a transaction their agent made, or a decision their AI recommended, and the answer to "who is responsible" will not be obvious.

The honest answer right now is that we have not figured it out. The technology is running a few steps ahead of the frameworks we would need to make sense of it.

That gap is worth watching. Not with worry, but with attention.