AI is making scams look official — and fake customers look real

Information Technology 4 min 70 sources

A wave of GTA 6 "beta test" scams and undisclosed AI influencers share one engine — the cost of looking authentic has collapsed. Plus a push for higher big-tech taxes and the next AI bottleneck.

Key takeaways

A GTA 6 "beta test" scam is stealing logins and bank details — AI now makes fake sites and emails look completely official, so polish is no longer proof of anything.
Brands are quietly using AI-generated "influencers" posing as real customers, with no UK rule requiring disclosure; 70% of people can't reliably spot fakes.
Two-thirds of Britons want higher taxes on big tech, and a new analysis says the real brake on AI at work is training people, not buying tools.

The cheapest thing to fake right now is the look of being real. Two separate stories this weekend — a scam targeting people waiting for a video game, and brands quietly using fake “customers” to sell things — are the same story underneath. AI has made a polished website, a fluent email, and a smiling human face nearly free to produce. The signals people use to tell real from fake are the exact things getting copied.

A fake game invite, a real bank-account risk

Millions of people have waited years for Grand Theft Auto VI, the next chapter of one of the best-selling video game series ever. It is due out on 19 November after two delays.[1] Criminals are using that wait against fans.

An email arrives inviting you to a “beta” — an early test version you play to help spot bugs before launch. It looks official. There is no such program. Rockstar Games, the maker, has announced no public beta test.[1]

The mechanism is simple and old; the polish is new. “With the help of AI, scammers can actually mimic official websites really, really well,” said Gerald Kasulis of the security company NordVPN.[1] One fake site reads: “We need you to help us build Vice City” — the city where the game is set — “we’re inviting a select group of players to experience the game early.”[1] Click through and you are asked for your name, address, date of birth, or your login for the existing online game. In one case, the “beta” was a download called GTA Mobile 6 that carried malware — software that let fraudsters connect to the victim’s computer and reach for bank details.[1]

The angle for anyone online: a polished site is no longer evidence of anything. The old rule — typos and bad design mean a scam — is dead. Treat unsolicited early-access offers as fake unless they come from the company’s own store or site. If you have already entered a gaming password into one, change it now.[1]

The customer who isn’t a person

The same week, a Guardian investigation found brands quietly using AI-generated “influencers” — fake people who post as if they were genuine, satisfied customers.[2]

In one set of videos, a bride cries with happiness about a photo app she used at her wedding. She isn’t real.[2] In another, a woman praises a home-design app called Maket; the company confirmed it had used AI influencers “to test creative concepts.”[2] A clothing brand called Ashle posted a photo of a “customer” with an extra finger — a classic AI giveaway — then deleted the images when asked about them.[2]

No specific rule in the UK requires brands to say when a person in an ad is AI-generated.[2] The EU’s AI Act starts requiring clear labels on deepfake content in August; that law does not apply in Britain.[2] The scale of the problem: a Which? investigation found 70% of people could not correctly sort real videos from fake ones.[2] “Consumers are not able to trust the content they are seeing online,” said Lisa Barber of Which? Tech.[2]

The thread between the two stories is the cost of authenticity. A face, a five-star testimonial, an official-looking page — these used to be expensive to fake, so seeing one was decent evidence it was real. AI made all three cheap. The signal still arrives; it just stopped meaning anything.

Britain wants big tech to pay more

A poll released Monday found 67% of Britons want the government to charge higher taxes on multinational tech firms like Meta, Google, and Amazon.[3]

The UK’s digital services tax — introduced in 2020 — is a 2% levy on the revenue of search, social media, and marketplace companies with UK sales above £25m (and £500m globally).[3] Only a handful of firms pay it. It raised about £800m last year.[3] The Fair Tax Foundation, which ran the survey, found two-thirds of respondents backed raising that rate.[3]

Why now: the digital services tax has been a recurring bargaining chip in trade talks with the US, which views it as a tax aimed at American companies. Public sentiment like this gives a government cover to hold the line — or raise the stakes.

The bottleneck nobody priced in

One quieter signal worth tracking: an analysis in the Financial Times argued that the next thing slowing AI’s rollout inside big companies isn’t chips or models — it’s people.[4] Tools are being bought faster than staff can be trained, trusted, or reorganised to use them. The expensive part of adopting AI is turning out to be the human part, which no purchase order fixes.[4]

If you work anywhere near AI rollout, that’s the number to watch — not how powerful the model is, but how long it takes a team to actually change how it works around it.

02 · Lesson · why it matters

The proof you trust is the first thing worth faking

We judge what's real by shortcuts — a polished site, a real face, a fluent email — and when those shortcuts get cheap to copy, the work of checking quietly lands back on you.

A door that looks like every other door

A fan opens an email inviting them to play a game they’ve waited years for. The site is clean. The wording is confident. There’s a code, a logo, the right city name. Everything that usually means “this is real” is there. None of it is.

For most of the internet’s life, that polish was a decent test. Real companies could afford clean design and fluent writing; scammers usually couldn’t. So you learned a shortcut: typos and ugly pages mean danger, smooth ones mean safe. The shortcut worked because faking it was expensive.

That’s the thing that changed. AI made the polish nearly free. The door still looks like every other door — it just no longer tells you what’s behind it.

Trust is a shortcut, not an inspection

Here’s the quiet truth under both of this week’s stories. Nobody verifies most of what they trust. You don’t read a company’s source code before logging in, or background-check the smiling person recommending a photo app, or trace where an email really came from. You can’t — there isn’t time. So you lean on signals that stand in for the real check: a face means a person, a polished page means a real business, a five-star testimonial means a happy customer.

This is not laziness. It’s how trust has to work at scale. A signal is only useful as a shortcut when it’s costly to fake — when producing the signal honestly is about as hard as faking it. A handwritten signature, a company’s clean storefront, a human face on a video: each used to take real effort to counterfeit, so each was worth believing.

The moment faking the signal gets cheaper than earning it, the signal stops carrying information. It still arrives. It just means nothing now.

What gets faked first is what you trusted most

Notice which signals AI went after. Not obscure ones — the load-bearing ones. The official-looking site. The human face. The fluent, confident sentence. The satisfied customer.

That isn’t a coincidence. The most valuable signal to fake is precisely the one people rely on most, because that’s where the payoff is. A scam works by borrowing trust you’ve already extended to something real — the look of Rockstar, the feel of a genuine review. The forger doesn’t invent a new kind of proof; they copy the proof you already accept.

So the better a shortcut works, the bigger the target it paints. The fake bride crying about her wedding app and the fake GTA beta key are doing the same job: wearing the costume of the exact thing you’d let your guard down for.

The cost didn’t vanish — it moved to you

When a signal stops working, the need to verify doesn’t disappear. It just stops being someone else’s job and becomes yours.

For years, the platform, the design, the polish did some of your checking for you. Now you have to do it by hand: go to the company’s own site instead of the link, change a password you typed into the wrong box, look twice at the customer who has six fingers. The work that used to be folded invisibly into “it looks official” is now a tax on your attention, paid every time you decide whether to believe something.

And it doesn’t land evenly. The people with the least time, the least practice spotting fakes, the most eagerness — the fan who’s waited years, the shopper who just wants the thing — are exactly the ones the scam is built for. The bill for cheap fakery falls heaviest on whoever can least afford the extra suspicion.

What seeing the whole leaves you holding

It’s tempting to read all this as “be more careful online,” and stop there. That’s the clever version. The humbler version is to notice how much of your daily trust runs on shortcuts you never chose and can’t really inspect — and that you are not standing above this problem, you’re inside it.

You trust the green padlock, the verified badge, the familiar logo, the voice on the phone that sounds like your bank. Most of the time you have to. The web of trust that lets you function — buy, log in, believe a stranger — is the same web that’s now cheaper to imitate. You can’t opt out of relying on signals; there’s no time to verify everything from scratch. You can only hold each one a little more loosely, and remember that the smoother something looks, the more reason someone had to make it look that way.

That’s not a reason to trust nothing. It’s a reason to know which of your shortcuts are doing the most work — and to keep a hand on the ones that, if faked, would cost you the most.

03 · Lab · your turn

Trust or Check

Rehearse judging polished messages and feel that the look is the cheapest thing to fake — only a self-verifying source proves what's real.

04 · Hope · carry this

The reason fakes are worth making is that the real thing — a stranger's honest word, trusted on sight — is one of the most valuable things people have ever built together. We've been forced to relearn how to spot the counterfeit many times before, and each time we have.