AI Uniti

The face selling the scam works for you.

Through 2025 and into this year, fraudsters have been running deepfaked news anchors and financial analysts as paid advertising for fake investment products. Not a generic synthetic face. A specific, trusted, employed one.

One reported syndicate ran AI-generated video of well-known business news anchors and stock analysts as paid ads on Meta platforms across an eighteen day window last July. In another case, an anchor at a major broadcaster found herself in a Reel promoting a trading app, with the ad claiming her employer had vetted it.

Look at where the damage lands. The person who loses money did not trust an advertisement. They trusted a face, and the credibility of that face was built, paid for and protected by the organisation that employs it.

That organisation was never involved, is usually the last to find out, and absorbs the reputational cost anyway.

This is not only a broadcast problem. It applies to any business whose brand runs on named people: partners at professional services firms, fund managers, university leaders, chief executives who present results. If a person's face is an asset on your balance sheet in all but name, it is also an attack surface, and almost nobody monitors it the way they monitor the corporate brand.

The other thing worth noticing: this was distributed as paid advertising. That means there is an account behind it, a payment method, a campaign structure and a targeting choice. Detecting that the video is synthetic is one half of the answer. Who is running it, and who they aimed it at, is the half that lets you do something.

Whose faces does your organisation depend on, and who is watching them?

#NarrativeThreatIntelligence #Deepfakes #BrandProtection

1 day ago | [YT] | 1

AI Uniti

On 20 February 2026, India's amended IT Rules took effect. Platforms now have two hours to take down an impersonation deepfake after a valid notice, and three hours for other unlawful synthetic content. The old window was thirty six.

That is the first hard clock a regulator has put on this problem, and most of the commentary has focused on the wrong part of it.

The obligation sits with the platform, not with you. But read the condition attached: the clock starts on a valid notice. Not a complaint. Not a screenshot emailed to a support address. A notice complete enough to act on.

That quietly moves the burden. If your organisation is the one being impersonated, how fast the content comes down is now largely a function of how good your notice is. Every hour you spend assembling it is an hour the platform's clock has not started.

So the operational question is not "can we detect deepfakes". It is narrower and more awkward: when we find one at nine in the morning, what time is the complete, evidenced notice actually lodged?

Most organisations cannot answer that, because the work is manual. Find every copy. Establish which accounts are posting them. Timestamp the spread. Document the harm. Package it in a form a trust and safety team can action without redoing the work themselves.

The same rules require intermediaries to keep forensic-grade records of every takedown. Regulators do not use that phrase loosely. It tells you the evidentiary standard this whole system is moving towards.

India is first here. It will not be last.

Book a 15-minute Signal by AI Uniti demo: aiuniti.com/signal

If a synthetic video of your chief executive surfaced this morning, what time would your notice be lodged?

#NarrativeThreatIntelligence #Deepfakes #Regulation

1 day ago | [YT] | 6

AI Uniti

250,000 subscribers. Thank you!! We are super grateful for every single one of you and hope you are enjoying the content and it's helping keep you all safe.

We make these because the shift from watching what is said to watching who is coordinating is the most useful idea we know, and it is still badly under-discussed outside a small circle of researchers. A quarter of a million of you disagreeing with that last part is a good problem to have.

What is coming: the rest of Behaviour Speaks Louder Than Content, more case studies pulled apart minute by minute, and more of the questions you leave in the comments answered on camera.

Tell us what you want next. The most requested topic in the replies gets made.

Find us elsewhere: LinkedIn [www.linkedin.com/company/aiuniti] · X [@AIUniti] · Instagram [@ai_uniti]

2 days ago | [YT] | 5

AI Uniti

Every other person on the video call was synthetic.

February 2024. A finance employee at an engineering firm joins a call with his chief financial officer and colleagues he recognises. He makes fifteen transfers to five accounts, about twenty five million US dollars. None of it has been recovered.

Two similar attempts on other companies failed that year. Look at what stopped them. Not detection software. A person who thought something felt wrong, and an executive who asked a question only the real chief executive could answer.

Here is the honest part. No platform can promise to catch every synthetic video, because generation is improving faster than detection.

But a deepfake of an executive is rarely one artefact. It is the video, the accounts pushing it, the timing, and the audience it is aimed at. Detection tells you the media is fake. It cannot tell you about the four hundred accounts carrying it.

Two free things to do this week: decide what your verification question is, and who is allowed to bypass it.

Episode 9, the series finale. Enjoy! next week explore the companies who have been worst hit and their stories,

#NarrativeThreatIntelligence #Deepfakes #ExecutiveRisk

2 days ago | [YT] | 7

AI Uniti

"We are aware of accounts falsely claiming to represent us."

You have read this notice. Banks publish it about cloned apps. Universities publish it about accounts selling admission letters. Telcos publish it about fake support profiles.

Every one is the right thing to do. Every one is also an admission.

Read what it actually says: someone built an account wearing our name, it reached our customers, we found out afterwards, and we are now asking the people who were targeted to be more careful than we were.

A warning on your own website reaches the audience that already trusts you. It never reaches the audience that was intercepted.

The alternative is not a better notice. It is finding the accounts while they are still assembling.

#NarrativeThreatIntelligence #BrandProtection #Impersonation

3 days ago | [YT] | 4

AI Uniti

Ask a board who owns disinformation risk and you get seven answers, or none.

That is not an org chart problem. It is the nature of the threat. One coordinated campaign hits financial integrity, legal exposure, operational resilience and executive credibility at the same moment, so it never lands cleanly on a single desk.

Seven people each own a piece of it:

The CISO, because it is an attack surface and behaves like one.
The CFO and Investor Relations, because market capitalisation moves on manufactured narratives.
The General Counsel, because evidence you cannot explain is evidence you cannot use.
The Chief Risk Officer, because it belongs on the risk register with a named owner, not in a comms plan.
The CEO, because executive-targeted campaigns threaten credibility and tenure.
The CCO, first to feel it and usually last to receive any intelligence about it.
Government Affairs, because policy narratives are manufactured the same way.

Everyone has a stake. Nobody has the mandate. So it sits unowned until it becomes an incident, at which point all seven are in the same room asking the same question at once.

The useful part is that seven owners do not need seven tools. The same behavioural evidence, gathered once, answers all seven questions. One investigation, seven decisions.

Episode 8 is two minutes and walks all seven. https://youtu.be/KKrmvlPYPwU

If you had to name one person accountable for narrative risk in your organisation today, who would it be?

hashtag#NarrativeThreatIntelligence hashtag#RiskManagement hashtag#BoardRisk

3 days ago | [YT] | 10

AI Uniti

Four countries. One pattern.

In the past fortnight: a telco whose chief executive became the target instead of the company. A bank whose cloned apps reach customers before its own fraud warnings do. A university with an account selling admission letters in the Vice-Chancellor's name. A consumer brand that has lost track of which dealer pages are genuinely its own.

Different languages. Different regulators. The same behaviour underneath.

It is economics, not coincidence. Accounts are cheap, coordination is cheap, and the playbook travels without needing to understand the market it lands in. The creative gets localised. The behaviour does not.

That is why we score how accounts behave, not what they post.

Link in bio: aiuniti.com/signal

#NarrativeThreatIntelligence #BrandProtection #Impersonation

3 days ago | [YT] | 2

AI Uniti

Six hours you never get back.

Coordinated campaigns do not start loud. They start organised. Accounts assemble, cadence tightens, the same phrasing moves across platforms, all while the volume is still too low for anything to look wrong.

Conventional monitoring waits for volume, so the alert always arrives after the story. Across 793,000 analysed coordinated campaign videos, the pattern is the same every time: a quiet assembly phase, then amplification.

Signal by AI Uniti reads the assembly phase, 6 to 12 hours before conventional monitoring calls it coordination. That window is where a response is still cheap.

Link in bio: aiuniti.com/signal

#NarrativeThreatIntelligence #BrandRisk #Disinformation

5 days ago | [YT] | 12

AI Uniti

Hi Friends, a new video just landed: The EU just made deepfakes illegal to hide. There's one problem.

As of this month, the EU AI Act's Article 50 is enforceable. AI-generated content must be labelled, synthetic interactions disclosed, fines up to EUR 15 million or 3% of global turnover.

It is a useful law built on a flawed assumption: that the people deceiving you will fill in the paperwork. The coordinated campaign impersonating your executives or pumping a stock will label nothing. Compliance is for the honest.

In two and a half minutes, this video covers what the law requires, why it cannot catch the actors it most wants to, and how behavioural detection closes that gap: timing, coordination, network shape. Signals that need no label.

Watch it now: https://youtu.be/vnUdOkb2NGY

Labels catch the honest. What catches the rest at your organisation?

hashtag#NarrativeThreatIntelligence hashtag#EUAIAct hashtag#Deepfakes

6 days ago | [YT] | 1

AI Uniti

183,000 of you, and more than 146,000 arrived in the last 28 days. Thank you so much for all of us. When we started making videos about narrative threat intelligence, we did not expect this. New here? Start with Episode 6: Is It Real, and Who Is Pushing It?

Thanks again to all of you!!

Signal by AI Uniti: aiuniti.com/signal

#NarrativeThreatIntelligence #YouTube #CommunityMilestone

1 week ago | [YT] | 11