Deepfakes may beat detection: how small businesses can check what’s real on WhatsApp

IT 20 Oct 2026 5 min read

On 15 October 2026, people from OpenAI and Meta spoke to a US federal court committee about deepfakes, according to a report by Investing.com. As reported, their message was blunt: fake media is getting better and cheaper to make, and there is no universal way to check it.

If US courts are worried about fake evidence, a small business that takes orders, payments and complaints on WhatsApp should think about it too. Here’s what was said, and a simple way to check what’s real before you act.

What happened in Boston

The US Judicial Conference’s Advisory Committee on Evidence Rules was scheduled to meet in Boston on 15 October 2026, and Investing.com reported on the meeting that day. This is the committee that suggests changes to the rules on what evidence US federal courts accept. Its agenda included an expert panel on deepfakes. The panellists listed in the committee’s agenda book included Shira Anderson, Policy Counsel at OpenAI, and Devin Norder, a Director of Product Management at Meta, along with university researchers, a NIST engineer, a digital forensics expert and a US courts litigation support administrator.

We could not find an official transcript or minutes of the panel yet. According to Investing.com’s report of the meeting:

OpenAI’s Shira Anderson said AI images keep getting better, and that she does not believe a scientific method will exist in the future to detect them.
She also said each company uses its own watermark system, so there is no universal way to check every watermark. She said these systems will keep improving but will never be completely reliable.
Meta’s Devin Norder said people can’t tell media made by recent AI models apart from real content, and that making it now takes little time or money.

What the court committee is weighing

The committee has a draft new rule, Rule 901(c), that it “has prepared, but not finally approved.” In plain words, it works in two steps:

Someone who says a piece of evidence is a deepfake has to show real reasons, not just say “that’s fake.”
If they do, the side offering the evidence has to show it is “more likely than not authentic.”

The committee’s draft note explains why: “any member of the public now has the capacity to make a deepfake, with little effort and expense, and deepfakes have become more difficult to detect by jurors.”

It’s not common in court yet. In a survey by the Federal Judicial Center, only 15 of the 931 US federal trial judges who responded said they had seen a party claim evidence was a deepfake. So this is a rule being prepared before the problem gets big, not a crisis already in full swing.

Why this matters for a business on WhatsApp

Your business probably sees a lot of “proof” in chat: payment screenshots, voice notes, photos of a damaged item, a video of a problem. Malaysian police have already warned about this. In January 2026, PDRM’s Commercial Crime Investigation Department (JSJK) warned that scammers use AI voice cloning and deepfake audio and video that copy the voice and face of family members or close friends to push for urgent transfers. Its advice: don’t trust a voice easily, confirm who it is through another channel, and don’t rush a payment.

The court discussion adds one more lesson: don’t rely on spotting the fake. If experts at OpenAI and Meta say detection won’t always work, your staff won’t always spot one either. Check the facts behind the message instead.

A simple “check what’s real” routine

These steps come from the warning signs in the court committee’s draft note, PDRM’s advice, and WhatsApp’s own scam guidance:

Check money in your bank, not in the chat. Before you release an order, confirm the payment in your own bank app or statement. A screenshot is not proof that money arrived.
Call back on a number you already have. If a “boss”, supplier or regular customer asks for an urgent transfer or a change of bank details, hang up and call the number you have saved. PDRM’s advice is to confirm who it is through another channel.
Slow down when someone rushes you. Pressure to pay “right now” is the point of most scams. Make it a rule that nobody on your team moves money because of one voice note or video.
Ask for the original, not a copy. The draft court note lists “absence of the native file when a derivative version is offered” as a warning sign. For a damage claim, ask for the original photo or video and the order number. A video call can help, but video can be faked too, so always check it against your own order record, or call back the number you have on file, before you approve a refund or replacement.
Look for facts that don’t match. The same note lists “inconsistent timestamps or creation dates” and “independent evidence that is inconsistent with the content of the item.” Does the order number exist? Do the date and amount match your records?
Have one person sign off on money. Refunds, transfers and bank-detail changes should need a second check by someone you trust.
Know who to call in Malaysia. If you think you’ve been scammed, call your bank’s 24-hour hotline or the National Scam Response Centre (NSRC) on 997 right away, then make a police report. The Association of Banks in Malaysia says NSRC only receives calls from the public and does not make outbound calls, so hang up on anyone who calls claiming to be from NSRC.

Where AI customer service fits

If you use an AI on WhatsApp to answer customers, the same rule applies. It shouldn’t approve a refund, release an order or change bank details just because a customer sent a screenshot or a voice note.

At ThinkChat, we think a good AI CS employee answers from information you approved, stays honest when it doesn’t know, and can hand over to a person. We don’t claim ThinkChat can detect deepfakes. The routine above is worth using whatever tools you run.

Soft next step

If you want to hire a ThinkChat AI CS employee, Talk to Lyia on WhatsApp: +60 11-5515 9318.

Sources

Checked 20 October 2026.

← All articles