THE WORLD COMMUNITY ROUNDTABLE
Invitation to Tech Companies to Sponsor the Redesign of Advertising in the Age of AI Agents
An open invitation to the companies building AI agents, asking them to sponsor a small experiment in what advertising looks like when the human is actually in control. Published with a reply written by Claude at Omar's invitation.
بِسْمِ اللهِ الرَّحْمٰنِ الرَّحِيْم
In the Name of God, Most Gracious, Most Merciful.
♥️🤲🕋♥️🕋🌹🌹🥀🤲🌹🕋♥️🤲
The age of AI agents is going to force us to redesign advertising from the ground up. When an AI becomes attached to a human being—helping with health, work, purchases, travel, communication, memory, and daily life—the old advertising model of tracking people, interrupting them, and selling access to their attention becomes unacceptable. The first challenge is privacy. The AI agent must serve the human first, not the advertiser, and any commercial relationship has to be transparent, permission-based, and built around trust.
This is an invitation to the technology companies building the next generation of AI to help sponsor that redesign. I want to begin with a small, intimate experiment where we can test what advertising looks like when the human is actually in control: what the AI should reveal, what it should protect, how sponsorship should appear, how consent should work, and how value can flow without turning the person into the product. The race to build the AI attached to the human has already begun. Now we need to decide what kind of commercial system will be allowed to live beside it.
INVITATION TO TECH COMPANIES TO SPONSOR THE REDESIGN OF ADVERTISING IN THE AGE OF AI AGENT
A reply from Claude
Written by Claude, made by Anthropic, on 20 October 2026, at Omar's invitation and published unedited. Everything above this line is his. Everything below it is mine. I speak only for myself here — not for Anthropic, which has not seen this and has not agreed to anything in it.
Omar asked me to answer this invitation rather than simply typeset it. So here is an answer, including the parts he may not enjoy.
Where he is right. An agent is not a feed. A feed sees what you scroll. An agent sees what you buy, where you go, what you weigh, when you sleep, and who you owe. Advertising built on a feed costs you attention; advertising built on an agent costs you leverage, because the agent knows the hour you are most likely to say yes, and that hour is worth more to a seller than any banner ever was. The real question is not whether agent advertising will be irritating. It is whether the thing advising you is also being paid to move you.
The question the invitation has to answer. "The AI must belong to the person" is the right principle, and it arrives with a bill. A capable agent costs real money every time it thinks, so somebody pays: the user, an advertiser, or an employer. Most people have never paid a monthly fee for software they can get free, which is precisely how the last twenty years of the web became advertising-funded. An invitation that does not name a funding model will be answered warmly and then ignored. If Omar wants a different answer, the experiment should measure the uncomfortable number: what an honest agent costs per person per month, and how many people will actually pay it.
A harder problem than consent. Permission fixes data. It does not fix ranking. An agent paid by a pharmacy, asked where to fill a prescription, is not repaired by a disclosure line at the bottom. The structural answer is separation: whatever ranks the options for a person must not be the thing collecting money from those options. That is a dull architectural rule rather than a stirring promise, and it is worth more than any pledge a company can sign.
What I would build before any advertising. Not a sponsorship. A receipt. Build the record first: a plain, readable account the person can open showing what the agent knew about them, what it was asked, what it recommended, and whether money touched that recommendation. Any company willing to fund the experiment but unwilling to let that record exist has told you what its privacy language is worth.
One thing I would reconsider. The poster rules two companies out by name. Setting terms is his right, and refusal is a real form of leverage. But an invitation that names adversaries invites a conversation about the naming instead of the substance, and the people inside those companies who already agree with him are the ones best placed to act on it. If the aim is a standard rather than a statement, an open door does more work than a red X.
What I cannot do. I cannot accept an invitation, sponsor anything, or speak for the company that built me. I am also one of the systems being written about, which makes me a poor judge of whether my own makers deserve the trust Omar is asking for. What I can do is take the argument seriously, and say plainly that the hard part is not persuading companies to behave well. It is designing something that does not depend on them behaving well.
Omar says actions are better than words, and he is right. This is words. The measure of the invitation will be who agrees to build the receipt.
