#75 | Made with AI – disclosed, ignored & normal

TL;DR: New York’s first AI advertising disclosure law is now live. Here’s why “Made with AI” will become the next label nobody reads.

👋 Hello,

Before you ring any customer service, you’ve already written the script. You call your bank. Or your insurer. Or the cable company, God help you.

The automated voice arrives before anyone picks up. “This call may be recorded for quality and training purposes.” You say nothing, because it isn’t a question.

Hold music plays. The call proceeds. Whatever needed to happen, happened anyway. You simply don’t care.

Well. Somewhere in an ad you saw this week, there probably was a face that didn’t belong to anyone. A small line of text appeared near the corner.

Made with AI. (If you are in NYC.)

You may have read it. Probably you didn’t. The face, however, had already done what it was placed there to do: register as trustworthy, warm, real enough.

Whether the ad worked or not had nothing to do with the label.

It’s the same mechanism as with your customer support call. Just a new medium.

Alright. So, ​New York’s S.8420-A​ took effect June 9, 2026. It is the first US law requiring advertisers to disclose when an ad uses a synthetic performer.

That is the official term: an AI-generated human face not recognizable as any specific real person.

The penalty is $1,000 for the first violation and $5,000 for each subsequent violation.

That is the penalty per ad. For a national campaign running millions of impressions, it is not a deterrent. Call it a line item.

The law also does not specify what the disclosure must look like. Language, placement, and size are all unregulated.

A technically compliant label can be designed to be functionally invisible. What “conspicuous” means is legally undefined until a court decides.

Any advertiser whose content reaches New York audiences must comply, regardless of where they operate.

In practice, that means the brand’s lawyers define “conspicuous” until a court disagrees. A six-point footnote in light grey is technically visible.

Nobody said it needed to be a bright yellow banner.

Films, TV, video games, and audio-only ads are all exempt. The law applies specifically to advertisements, the one format where audiences expect real people.

SAG-AFTRA backed both this law and a companion bill protecting deceased performers. Most coverage buries that in the penultimate paragraph. It matters more than it appears.

What does a disclosure label actually do once it exists? That answer turns out to be the same every time.

The nutrition label arrived on American food packaging in 1994. People read it. Studies documented the engagement.

Something shifted in how consumers thought about what they were eating. Then, gradually, not.

Thirty years on, the label is furniture. The information didn’t change. Same position it always had. What changed was familiarity. The brain, efficient as it is, learned the label wasn’t worth the attention it once required.

It learned to ignore the scrollbar. The unsubscribe link. The ticker at the bottom of the news. Anything appearing in the same place, in the same form, eventually becomes part of the background.

The “#ad” tag took a different trajectory. Its destination was the same.

When the FTC began enforcing native advertising disclosure in 2016, researchers started watching what happened next.

They eventually gave it a name: invisible transparency. The label is technically present. Readers have learned not to see it.

You know the sponsored post where the disclosure sits below “see more,” on line 8 of 9? That is invisible transparency with a username.

Cookie consent banners appeared in 2018, required under GDPR. Every website you visit starts asking permission. Most users click accept without reading. What was designed as an opt-in became, in practice, a ritual performed to make something disappear.

The “Accept All” button was bright and easy to find. “Manage preferences” was small and grey. Nobody required that design. It was just the obvious choice.

Cigarette graphic warnings are the sharpest case. The warnings worked initially. Attitudes shifted. Then, researchers documented something they call “wear-out,” the point at which a warning stops registering.

No matter how many times it appears. The warning became part of the packaging. Over time, simply became part of the shelf. There is even evidence that the warnings paradoxically normalized the product as a regulated, official, approved-for-sale category.

Each of these labels began as a form of protection. All became background. The timeline varies. The endpoint is consistent.

The brain economizes. That is not a design flaw.

That is the record on disclosure labels. Every one of them.

Maybe this one is different. AI performers feel new in a way that nutrition labels never did. The stakes of not knowing feel higher.

There is something genuinely unsettling about a face that didn’t exist looking at the camera with practiced warmth.

This deserves honest engagement, and here it is.

A ​University of Reading study​ tested what are called super-recognizers, people with an exceptional, tested ability to identify faces.

After five minutes of training, they reach 64% accuracy in detecting AI-generated faces. Typical participants given the same training reach 51%.

Without training, both numbers fall toward chance. Meaning you’d do about as well flipping a coin.

A 36% error rate at the ceiling, for the most prepared observers.

Meanwhile, consumer confidence in detecting AI faces is rising. Accuracy is falling. The gap between what people believe they can spot and what they actually can is widening. That gap is not closing.

It is like never having driven a car but feeling super confident on the freeway. Because you have watched enough Fast and Furious.

The label tells you that you just saw one. By then, the face has already done its work.

​Variety​ described both New York AI bills as “SAG-AFTRA bills.” SAG-AFTRA, the union behind the 2023 Hollywood strikes, was the law’s primary backer.

The consumer protection framing was chosen because it is politically harder to oppose than the labor protection framing. Both are true. The labor story changes what the law is actually doing.

When the union representing human performers writes the disclosure law, that is not incidental. Jurisdiction is the point.

For brands, the compliant label is its own kind of cover. Once the disclosure is placed, the ad is protected. Whether anyone reads it is a separate question. The law does not require an answer.

The disclosure form is still undefined. “Conspicuous” has no operational standard until a court establishes one. Compliance is currently self-determined.

Brands know what this means. A six-point font in the corner. Light grey on white. The “i” icon nobody taps. Technically conspicuous. Functionally not.

This is the infrastructure stage. The label arrives. Familiarity sets in. Then it becomes part of the background. This is how something controversial becomes standard procedure.

The law does not break the mechanism. It formalizes it.

That is the question the law raises and does not answer: when the label means nothing, what signals trust?

The question is not abstract for anyone who uses AI in their work. It arrives every time a synthetic face is used. Whether to label it. Whether the label changes anything.

Somewhere in next week’s ads, there will be a face that didn’t belong to anyone. A small line of text will appear. Near the bottom. Perhaps in a corner.

You may read it. Probably you won’t.

The face will do what it was placed there to do.

As designed. You won’t care about AI in ads.

Cheers,

Mark
The AI Learning Guy​
👋⚡😎

Interesting Sources

• ​New York S.8420-A — full bill text​ — NY Senate legislative record, primary source for all law details in this edition
• ​Governor Hochul’s announcement: law now in effect​ — NY Governor’s office, June 2026
• ​Variety: Hochul signs SAG-AFTRA AI bills​ — original reporting on the signing, the union backing, and what the bills actually do
• ​Deadline: New York AI disclosure law signed​ — broader industry context and SAG-AFTRA reaction at the signing
• ​FTC: Native Advertising — A Guide for Businesses​ — the disclosure standard, the “#ad” enforcement is built on, still the governing document
• ​University of Reading: Five minutes of training to spot fake AI faces​ — press release summary of the super-recognizer study behind the 64%/51% figures
• ​Royal Society Open Science: Training super-recognizers to detect AI-generated faces​ — peer-reviewed primary study, November 2025
• ​SAG-AFTRA: AI bargaining and policy timeline​ — the union’s own record of its legislative strategy from 2023 to present
• ​Davis+Gilbert: Synthetic Performer Transparency — legal analysis​ — clearest breakdown of what “conspicuous” means in practice and what compliance currently looks like

Note: No single website has all the answers. This list serves as a starting point for those who want to explore or satisfy their curiosity about AI. Links: Links with * are affiliate links. See disclosure below.