Protecting the freedom to think

Our own infrastructure must face the same scrutiny we turn on AI detectors and disclosure rules. Making human judgment recognizable is meant to build trust. It can also become a new way to shut people out. If receipts became a badge everyone was expected to wear, anyone without one would start to look as if they had something to hide. Any way of recognizing judgment can miss judgment that works differently.

JA Westenberg’s video on the AI purity test shows what this kind of suspicion costs writers, and why “certified human” badges hurt those who most need protection. One viewer wrote:

I’m a beginner writer focusing on highly stylized short fiction who also happens to be neurodivergent and is writing in a second language. AI detectors consistently flag my writing as ‘probably AI generated’. For a while, I even started actively changing my voice to not sound like AI, even though I wasn’t using it. Eventually I lost almost all motivation to continue learning how to write. I thought: What’s the point in making all of this effort if people are just going to assume I’m a fraud, all because they trust more in the opinion of an inherently flawed computer model and their own misguided gut-feelings. Toxic AI detection culture very nearly made me quit my biggest passion.

A receipt makes one narrow claim: that a particular piece of work met particular conditions. It does not vouch for anyone’s intelligence, honesty, or humanity. Where no receipt is required, lacking one should count against no one, and no system built on the protocol should flag its absence. Where a receipt is required, a missing one means only that a condition hasn’t been checked.

An institution that imposes conditions owes people a reason for them and a way to push back. People should be free to explore and change their minds without having to produce a record of it. Keeping exploratory conversations confidential is essential, but it is not enough.

People show judgment in different ways, so there should be many ways to show it. Our authorship check is one. A writer might share a moment when AI plainly missed the point and explain why. Another might spend an hour answering an audience’s questions about the work. A novelist might invite readers into parts of her process as a chef opens a kitchen, letting them watch, taste, and make suggestions. Communities can invent, test, and argue over many more approaches, matched to the purpose and stakes. Where the purpose allows, people should choose how to make their judgment recognizable. What one person freely shares should not become what everyone is expected to disclose.

Public rules and accountable governance

The infrastructure must itself be accountable to the people whose exchanges it governs. Three further conditions, developed in our Society article, follow:

• Open. What each receipt claims, how that claim is checked, and its limits should be publicly specified and open to challenge. Changes to those definitions must be explicit. Openness applies to the rules; it does not require exposing private exploration.
• Interoperable. Receipts should remain understandable and checkable across participating services. Changing providers should not change their meaning. Portability does not expand a receipt’s authority beyond the conditions it certifies.
• Independently governed. No single company, institution, or lab should control the shared framework. Teachers and students should help develop, test, and revise educational standards together. Students need a role in deciding which conditions are adopted, as well as a way to challenge them afterward. Across professional settings, those whose work the standards govern should participate in governing them.

These conditions make it possible to question the rules, choose among implementations, and improve the infrastructure without depending on one provider’s permission.

Any institution that requires verification should make the requirement simple to meet, at no extra cost to participants. Receipts should come from simply doing the work, in activities whose demands are justified by what the work is for.

We have to learn which approaches support judgment, protect freedom, and let more people take part, and which ones create new pressure to perform credibility. That question includes our own public demo, which lets anyone attach an “AI-free” receipt to what they write there. If it starts to work like a purity test, we want to know.