“Those photos are AI”: what happens when a tenant says it at the hearing

The first article on this site made one point: “those photos are AI” now costs a tenant nothing to say. This one follows that sentence into the hearing. What does a judge do with it? What did the people who write the federal rules of evidence decide about it this year? And what should you bring so the question has an answer?

This is not legal advice. Rules differ by state and by court, and a lawyer in your state is worth more than anything written here.

Small claims runs on who the judge believes

If a deposit dispute is heard in small claims court, the formal rules of evidence usually give way to judgement. Utah’s rule, for example, says the judge “may receive the type of evidence commonly relied upon by reasonably prudent persons in the conduct of their business affairs”, and the rules of evidence are not applied strictly. Check the rule in your own state.

That cuts both ways. Your photos are unlikely to be thrown out on a technicality, and neither is the tenant’s claim that they are fake. There is no expert and no lab. The judge hears you both and decides whom to believe.

What the federal rule-makers decided this year

The committee that drafts the Federal Rules of Evidence has spent three years studying AI. Small-claims courts do not use the federal rules, but its May 2026 report is the most careful thinking any court body has published on exactly your problem, and the part on AI is only a few pages. Three things in it matter to a manager.

• No new rule yet. The committee concluded that “at least for now, an amendment to Rule 901 to address deepfakes is not warranted”, because few courts have actually faced one. In a Federal Judicial Center survey of federal judges, “only fifteen respondents reported having dealt with deepfake issues.”
• A bare claim should not be enough. The committee keeps a draft rule ready in case that changes. Its first principle is that someone should not get an inquiry into whether an item is fake “merely by claiming that it is a deepfake”. They should have to show some reason to think so.
• Once there is a reason, “I was there” may not be enough. The second principle: if the other side does show a real reason for doubt, the person offering the photo must prove it is authentic more likely than not. That is a higher bar than usual. The draft’s note explains why a witness is not enough on their own: “a person with ‘knowledge’ may well think that the item is genuine when in fact it is a deepfake.”

Read those two principles from your side of the table. The first protects you from the free sentence. The second tells you what happens if the tenant brings more than a sentence. Your word is exactly what the committee says is not enough, so you need something besides it.

What “something besides your word” looks like

Things that exist outside your memory of the day:

• The original file, with the camera’s own data in it. The phone’s model, the exposure, the time on the camera clock. That is what the phone wrote, not what you say. Getting it off the phone without losing that data is covered in getting the original photo.
• A record made at the time. A photo you can show was on file three days after move-out is harder to call a fabrication than one that first appears the week before the hearing. The move-out email that carried the photos, or the inspection report you sent the tenant with them attached, is that kind of record.
• A fingerprint showing nothing changed since. A SHA-256 hash is a number computed from a file’s exact bytes. Change one pixel or one line of its data and the number changes completely. Federal courts already lean on this idea: Rule 902(14) covers data authenticated by “a process of digital identification”, and its committee note says that when the hash of an original and a copy match, “it is highly improbable that the original and copy are not identical.” In federal court that route also needs a certificate from a qualified person. The point for you is the idea, not the paperwork.

What to bring

For each charge the tenant disputes:

1. The photos as files, on a laptop or a USB stick. A print has no camera data in it. Bring prints for the judge as well, but the file is the thing that can be checked.
2. A list of the files. For each one: the camera model and camera time read from the file, the date you received or filed it, and its SHA-256.
3. Whatever shows when you had them. The email that carried them, the inspection report sent to the tenant.
4. The move-in photos of the same spots, listed the same way.
5. The tenant’s photos, if they sent any, listed the same way. If theirs read Can’t Verify, that is not evidence they are fake, and you should say so if asked.

You can make the list in item 2 yourself. On Windows, certutil -hashfile photo.jpg SHA256 prints a file’s SHA-256; on a Mac, shasum -a 256 photo.jpg does the same. Our checker reads the camera data for free. What a list you made yourself cannot show is when you made it. That is what our paid report adds: the same list filed with someone other than you on a date, at a web address the judge or the tenant can open and check against the files without trusting either of us.

Keep one copy untouched

Any change to a file changes its fingerprint. That includes cropping, drawing an arrow on it, rotating it and saving, and sometimes just re-saving it. Mark up a copy for the judge if it helps. Keep the original exactly as it came off the phone, and put that one on the list.

What this does not do

A matching hash is not a true photo. It shows the file is the same one listed on the date the list was made. It says nothing about whether the scene was honest, or whether the file was already fake when it was listed.

Camera data can be edited by anyone who knows how. It is strong corroboration, not a signature.

A stripped file is not a lie, on either side.

We cannot tell you how your judge will rule. We looked for published decisions on “it’s AI” claims in deposit cases and found none we could cite. If you have been through one, we would like to hear what happened: [email protected].

It does not give you a confidence score. We tried twice to build a classifier that judges a picture by its pixels. The second attempt looked at eighteen ordinary photographs and called seventeen of them fake. It is switched off. A file you can check is worth more in a hearing than any guess about one you cannot.

---

ImposterShield reads the evidence inside image files. It runs in your browser and uploads nothing. If a file has been stripped, it says so. That is the product working, not failing.