Videos showing famous people and government officials personally endorsing an investment platform, crypto giveaway, or miracle product are genuine recordings of those endorsements
By The Conspiratory EditorsAugust 2, 2026
Where the evidence lands: ContradictedThat an unsolicited video circulating in advertising or social media, in which a celebrity, executive, journalist, or government official appears to personally endorse a trading platform, cryptocurrency scheme, giveaway, or product with guaranteed high returns, is an authentic recording of that person making that endorsement.
Contradicted. This file rates a claim people encounter as an advertisement rather than as a conspiracy theory: that a video of a well-known figure endorsing a trading platform, a crypto scheme, or a health product is a real endorsement. In the documented cases it is not. The FBI's Internet Crime Complaint Center warned in December 2024 that criminals use generative AI to create images of celebrities promoting products and to produce videos for what it calls fictitious or misleading promotional materials for investment fraud schemes, and by July 2026 the same office was warning that fraudsters were circulating an AI-generated video of a senior FBI leader directing people to a spoofed IC3 website, in order to re-target people who had already been defrauded. India's Press Information Bureau debunked a video circulating in paid social advertising that appeared to show Finance Minister Nirmala Sitharaman endorsing an investment platform; she had endorsed nothing. Keep two things apart. Public figures do make real endorsements, and this file does not suggest otherwise. The rated claim is that these specific unsolicited videos, offering improbable returns and routing you to a deposit page, are genuine. They are fabrications, and they are the point of contact where synthetic media stops being an argument about truth and starts taking money.
Believed by: Not a movement and not an ideology. The victims are ordinary people scrolling past a paid advertisement, skewing older but by no means exclusively so, who have no reason to expect that a moving, speaking video of a recognizable person is manufactured. Belief here is not commitment to a theory; it is the ordinary assumption that a face on video belongs to the person it looks like.
Opening entry. Two threads are worth watching as this file develops. The first is whether provenance signals reach this category at all: OpenAI became C2PA-conformant and began embedding SynthID watermarks in May 2026, but a fraudster running a model outside those programs leaves no signal to find, so the coverage gap matters more here than in political deepfakes. The second is the recovery-fraud escalation in IC3's July 2026 alert, where an AI-generated video of a senior FBI leader steers people who have already been defrauded toward a spoofed complaint site. If that pattern spreads to other national fraud-reporting bodies, it will be the clearest sign the technique has industrialized. source →
The full story
The one deepfake story that ends in a bank transfer
Almost everything written about synthetic media concerns belief: whether a clip will swing an election, whether people can still tell what is real. This is the branch where the question is narrower and the answer is measurable. Somebody sees a video, believes it, and sends money.
The construction is consistent. A recognizable person, a film star, a technology executive, a finance minister, a familiar broadcaster, appears to describe an opportunity: a platform, a limited window, returns that are large and framed as low-risk. The video is bought into someone's feed as an advertisement. Clicking leads to a news-styled article about the endorsement, then to a platform, then to a deposit page.
What the FBI put in writing
On 3 December 2024the FBI's Internet Crime Complaint Center published announcement I-120324-PSA, and its value is that it is an inventory rather than a warning. It sorts the toolkit into text, images, audio and video, and names the uses.
Under images: criminals use generative AI tools to create images of celebrities or social media personas promoting counterfeit products. Under video: to create videos for fictitious or misleading promotional materials for investment fraud schemes. Under audio: to impersonate well-known public figures to elicit payments. Under text: to generate content for fraudulent websites for cryptocurrency investment fraud, with AI chatbots embedded in them.
Read together, that last item is the one people underestimate. The video is not the fraud; it is the doorway. Everything behind the door is generated too, which is why a skeptical victim who goes looking for confirmation finds some.
One case, run twice, numbers changed
India's Press Information Bureau Fact Check unit flagged a video in July 2026 that appeared to show Finance Minister Nirmala Sitharaman endorsing an investment platform. PIB said the clip was AI-generated and represented no communication from the minister or the government.
Two details make it instructive. First, it behaved as a template rather than a single artifact: versions promised anywhere from 2.5 lakh to 25.5 lakh rupees a month on the same 22,000-rupee stake, with the figures swapped between uploads while the footage stayed put. Second, it recurs. Fabricated endorsements attributed to this same minister had circulated and been publicly denied before this round appeared.
A campaign that can regenerate its creative and re-enter paid distribution after a national government has publicly debunked it is not really being fought on the question of authenticity. It is being fought on distribution economics.
Why 'count the fingers' is expiring
The FBI's own tips include looking for distorted hands and feet, unrealistic teeth and eyes, inaccurate shadows and lag time. That advice was sound when written and is worth a moment's attention. It is also depreciating fast.
Jeff Hancock, who founded the Stanford Social Media Lab, told NBC News in January 2026 that assessing an image or video by eye will essentially become impossible, and that the old AI literacy ideas about counting fingers are likely to go away. Any defense that depends on the generator being imperfect has a shelf life set by somebody else.
What does not expire is external verification, and it is oddly simple. Has the person said this anywhere they control? Does any news organization report the endorsement? Is the entity regulated where you live? Did this reach you unsolicited, and does it want money now? None of those questions require you to see anything the forger got wrong.
When the fraud impersonates the fraud desk
On 20 July 2026 IC3 issued alert I-072026-PSA, updating an April 2025 warning, and it describes a genuinely bleak escalation. Scammers were impersonating FBI personnel to approach people who had already lost money, almost always claiming to have recovered their funds or offering to help recover them.
In one variant an AI-generated video of a senior FBI leader urged viewers to file complaints on a spoofed IC3 website. The fake site copies the real one's look, but every link returns you to the home page, and the complaint form collects only what a fraudster needs: name, phone number, email address, scam type, and estimated loss. Then it issues a reference number and promises someone will be in touch.
The victim pool is a list of people who have already demonstrated they can be reached. IC3's answer is worth memorizing: it maintains no social media presence, and it does not investigate crimes or recover funds through Facebook, Telegram or similar platforms.
Read the offer, not the pixels
Keep the two things apart. Public figures do endorse products, and some of them endorse financial products; nothing here says otherwise. What this file rates is the unsolicited video promising improbable returns and routing you to a deposit page, and on that the answer is that these are fabrications.
The useful shift is where you look. Almost all public advice about deepfakes trains people to examine the image, and the image is the part getting better every quarter. The offer is not improving. A guaranteed return, a closing window, a platform you have never heard of, an approach you did not ask for: that combination was a fraud before generative video existed and remains one afterwards. The technology changed who appears to be talking. It did not change what they are asking you to do.
What's still unexplained
- The scale of the losses is genuinely unclear. National regulators publish warnings more readily than reliable aggregate figures, and much of the circulating loss data traces back to low-quality outlets recycling one another. That this is happening is well established; how much it costs is not.
- Whether provenance signals will help here is unresolved. Watermarking and content credentials are designed for the platforms and models that adopt them, and a fraudster running an open-weight model outside those programs leaves nothing to find.
- Legal responsibility is unsettled. Where liability sits between the advertiser, the ad platform, the model provider, and the payment processor differs by jurisdiction and is being litigated and legislated in several countries at once.
Point by point
The claim: It is clearly them. You can see their face and hear their voice, so the endorsement must be real.
What the record shows: A face and a voice are exactly what these tools produce, and they are now the cheapest part of the operation. The FBI's December 2024 announcement lists the components separately because they are assembled separately: generated video of a public figure, cloned audio, generated stills for supporting profiles, and generated text for the fraudulent website the video points to. Recognising someone is no longer evidence you are looking at a recording of them. The verification that still works is external: does the person, or their verified account, or any news outlet, say they did this?
The claim: If it were fake, the platform hosting the advertisement would have taken it down.
What the record shows: Enforcement runs behind distribution by design. These campaigns are bought as advertising, rotate creative and landing pages faster than review cycles, and often run under recently created accounts. The Sitharaman video is the illustrative case: it circulated in paid placements, was debunked by the government's own fact-check unit, and kept reappearing with the numbers changed. That a fabrication is being served to you as an ad is a statement about ad review, not about authenticity.
The claim: The old advice works. Look for the distorted hands, the wrong number of fingers, the smeared background.
What the record shows: That advice is in the FBI's own tip list, alongside irregular teeth and eyes, inaccurate shadows and lag time, and it is worth trying. But it is a depreciating asset. Jeff Hancock of the Stanford Social Media Lab, quoted by NBC News in January 2026, put it bluntly: judging an image or video by eye is approaching impossible, and the old AI-literacy heuristics about counting fingers are on their way out. Detection by inspection is a race the viewer loses. The durable check is not visual at all.
The claim: This is just ordinary scam advertising with a new coat of paint, not worth a case file.
What the record shows: The difference is whose authority is being spent. A generic scam ad borrows nothing; a fabricated endorsement borrows the credibility of a named person or institution, and it damages them whether or not anyone is defrauded. That is why the debunks increasingly come from the impersonated party itself, a finance ministry, a regulator, a broadcaster. It also connects this to the broader problem: every convincing fake endorsement makes it marginally easier for a real recorded statement to be dismissed as fake, which is the mechanism described in this site's file on the liar's dividend.
The claim: Nobody would fall for a video promising huge guaranteed returns.
What the record shows: The offer is not aimed at everyone; it is aimed at whoever the targeting delivers, and it only needs a small response rate to pay for itself. The unlikely return is doing deliberate work, filtering for people willing to entertain it, which is a long-established property of fraud rather than a flaw in it. And the follow-on announcement from IC3 in July 2026 documents the crueller stage: fraudsters using deepfaked officials to approach people who have already been defrauded once, offering to recover their money.
Timeline
- 2018-2022The template exists before the technology is good. Fake celebrity crypto giveaways spread using still photographs, doctored screenshots, and hijacked accounts, most visibly through impersonations of well-known technology executives promising to double any cryptocurrency sent to an address. The social engineering is already complete; only the video is missing.
- 2023Consumer-grade voice cloning and face-swapping become cheap and fast enough to run at advertising scale. Fabricated endorsement clips begin appearing in paid social placements rather than only in organic posts, which matters: paid distribution puts them in front of people who never sought them out.
- 2024-12-03The FBI's Internet Crime Complaint Center issues public service announcement I-120324-PSA, 'Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud'. It sets out the toolkit in categories: AI-generated text for fictitious profiles and translated approaches, AI-generated images including 'images of celebrities or social media personas promoting counterfeit products', cloned audio impersonating public figures and relatives, and AI-generated video used to create 'fictitious or misleading promotional materials for investment fraud schemes'.
- 2024-12-03The same announcement notes a detail that shapes everything downstream: criminals also 'generate content for fraudulent websites for cryptocurrency investment fraud' and embed AI chatbots in them. The video is the hook, but the whole apparatus behind it, the news-styled article, the platform, the support agent, is generated too.
- 2025-2026The technique globalizes. Regulators and fact-checking units in multiple countries issue warnings about fabricated endorsements attributed to national figures, financial journalists, and broadcasters. Reporting on aggregate losses is uneven and much of it circulates through low-quality outlets, so this file does not put a global figure on the harm; the official warnings are the firmer evidence.
- 2026-07India's Press Information Bureau Fact Check unit flags a video circulating in social advertising, largely on Facebook, that appears to show Finance Minister Nirmala Sitharaman endorsing an investment platform. PIB states the video is AI-generated and represents no communication from the minister or the government. Versions of the advertisement promise returns ranging from 2.5 lakh to 25.5 lakh rupees a month on the same 22,000-rupee stake, the figures swapped between uploads while the footage stays put.
- 2026-07-20The scheme turns on the people investigating it. IC3 issues alert I-072026-PSA, an update to an April 2025 warning, describing scammers impersonating FBI personnel to revictimize people who have already lost money. One variant uses an AI-generated video of a senior FBI leader urging viewers to file complaints on a spoofed IC3 site, which mimics the real one but collects only a name, phone number, email address, scam type and loss figure. IC3's answer is flat: it maintains no social media presence, and it does not investigate crimes or recover funds through Facebook, Telegram, or similar platforms.
- 2026-08State of the record. No fabricated endorsement examined by a fact-checking body or regulator has turned out to be genuine. The people depicted consistently deny involvement, and in several cases the depicted institution is the one issuing the debunk.
From the case file
The actual records: declassified, released, or leaked. We link straight to each document in its official archive, so you never have to take our word for it. Read the originals yourself.
PSA I-120324-PSA: Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud
The primary document for this file. Rather than a general warning, it is an itemized inventory of how generative AI is used in fraud, broken into text, image, audio and video, and it names the celebrity-endorsement and investment-promotion uses explicitly. It also carries the FBI's own detection tips and the advice to agree a verification word with family.
Read the document: IC3.gov →PSA: FBI Warns of Scammers Impersonating the IC3
The follow-on advisory, updating an April 2025 alert, documenting an AI-generated video of a senior FBI leader used to steer people to a spoofed IC3 site. Contains the statement most useful to a reader in that position: IC3 maintains no social media presence and does not investigate crimes or recover funds through Facebook, Telegram, or similar platforms.
Read the document: IC3.gov →Contradicted. This file rates a claim people encounter as an advertisement rather than as a conspiracy theory: that a video of a well-known figure endorsing a trading platform, a crypto scheme, or a health product is a real endorsement. In the documented cases it is not. The FBI's Internet Crime Complaint Center warned in December 2024 that criminals use generative AI to create images of celebrities promoting products and to produce videos for what it calls fictitious or misleading promotional materials for investment fraud schemes, and by July 2026 the same office was warning that fraudsters were circulating an AI-generated video of a senior FBI leader directing people to a spoofed IC3 website, in order to re-target people who had already been defrauded. India's Press Information Bureau debunked a video circulating in paid social advertising that appeared to show Finance Minister Nirmala Sitharaman endorsing an investment platform; she had endorsed nothing. Keep two things apart. Public figures do make real endorsements, and this file does not suggest otherwise. The rated claim is that these specific unsolicited videos, offering improbable returns and routing you to a deposit page, are genuine. They are fabrications, and they are the point of contact where synthetic media stops being an argument about truth and starts taking money.
Reviewed by The Conspiratory Editors · Last reviewed August 2, 2026 · How we rate
Common questions
Is AI endorsement scams true?
Contradicted. This file rates a claim people encounter as an advertisement rather than as a conspiracy theory: that a video of a well-known figure endorsing a trading platform, a crypto scheme, or a health product is a real endorsement. In the documented cases it is not. The FBI's Internet Crime Complaint Center warned in December 2024 that criminals use generative AI to create images of celebrities promoting products and to produce videos for what it calls fictitious or misleading promotional materials for investment fraud schemes, and by July 2026 the same office was warning that fraudsters were circulating an AI-generated video of a senior FBI leader directing people to a spoofed IC3 website, in order to re-target people who had already been defrauded. India's Press Information Bureau debunked a video circulating in paid social advertising that appeared to show Finance Minister Nirmala Sitharaman endorsing an investment platform; she had endorsed nothing. Keep two things apart. Public figures do make real endorsements, and this file does not suggest otherwise. The rated claim is that these specific unsolicited videos, offering improbable returns and routing you to a deposit page, are genuine. They are fabrications, and they are the point of contact where synthetic media stops being an argument about truth and starts taking money.
What is AI endorsement scams?
Most synthetic-media stories are about politics. This one is about money. Since roughly 2023 the same tools that generate political deepfakes have been aimed at a simpler goal: producing a video of someone recognizable, a film star, a business figure, a finance minister, appearing to endorse an investment platform, then buying advertisin…
What does the evidence show?
A face and a voice are exactly what these tools produce, and they are now the cheapest part of the operation. The FBI's December 2024 announcement lists the components separately because they are assembled separately: generated video of a public figure, cloned audio, generated stills for supporting profiles, and generated text for the fr…
Why do people believe it?
Moving video with matching audio has been reliable evidence for a century, and that expectation is not irrational, it is simply out of date. Most people have had no occasion to update it.
What is still unresolved?
The scale of the losses is genuinely unclear. National regulators publish warnings more readily than reliable aggregate figures, and much of the circulating loss data traces back to low-quality outlets recycling one another. That this is happening is well established; how much it costs is not.
Sources
- 1.Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud, FBI Internet Crime Complaint Center (IC3) (2024)
- 2.FBI Warns of Scammers Impersonating the IC3, FBI Internet Crime Complaint Center (IC3) (2026)
- 3.Viral AI video of India's FM Sitharaman offering huge returns is fake, Gulf News (2026)
- 4.AI is intensifying a 'collapse' of trust online, experts say, NBC News (2026)
- 5.FBI Warns of Deepfake Videos Impersonating IC3, Infosecurity Magazine (2026)
Embed this verdict on your site
Paste this snippet to show our sourced verdict as a small card, with a link back to the full case file. Free to use.
<iframe src="https://theconspiratory.com/embed/ai-endorsement-scams" title="The Conspiratory verdict" width="520" height="190" style="border:0;max-width:100%" loading="lazy"></iframe>Help us investigate
This is a living case file. If you spot an error or know evidence we missed, tell us, and weigh in on where you land.
Where do you land?
Cast your read on this one.
Comments
Add your take. Comments are read and approved by a human before they appear, so keep it on topic and civil. Please do not accuse named, living people of crimes.