Swedish households have lost more than SEK 500 million this year to AI-driven investment scams, according to new figures from the Swedish Economic Crime Authority (Ekobrottsmyndigheten) and the Financial Supervisory Authority (Finansinspektionen, FI). What used to be crude phishing attempts has evolved into a mass-production line of deep-fake endorsements, engineered with the same generative tools that power legitimate marketing.
The scale of the fraud
Between January and September 2025, FI logged 5,100 reports of investment scams—a 340 percent increase compared with the whole of 2022. The median loss per victim: SEK 98,000. The largest single loss reached SEK 4.3 million.
Most of those hit were first-time investors, with 78 percent trying equity or crypto products for the first time. Nearly two-thirds were over 55.
The AI playbook
Here’s how the scam works.
1. Scrape and clone. Criminals download hours of interview footage featuring Swedish public figures—actors, athletes, TV hosts—then feed it into open-source diffusion models. The result: convincing 30-second clips of these “celebrities” enthusiastically recommending an investment app.
2. Micro-target. The fake clips are distributed through Meta, TikTok, and Snapchat using look-alike audiences built around “cash-rich, finance-curious 50+” demographics.
3. Lock-in. Once the victim signs up, they’re taken to slick trading platforms that simulate real market movements. Prices rise, balances grow—but when the user tries to withdraw, they’re hit with “verification fees” or “tax clearance charges” that often exceed their balance. Then the account goes dark.
The regulatory blind spot
Under Swedish marketing law (Marknadsföringslagen 2008:486), the advertiser—not the platform—is liable for false or misleading promotions. When the “advertiser” is a shell company registered in the British Virgin Islands, FI has no one to penalize.
“We can issue a ‘stop-this-ad’ order, but the campaign is re-uploaded within hours under a new domain,” says Jimmy Kvarnström, FI’s Director of Market Conduct.
What’s about to change
Help may be on the horizon, though not immediately.
- The EU Digital Services Act (DSA) shifts enforcement to Sweden’s Consumer Agency (Konsumentverket) on 1 January 2026. From then on, major platforms could face fines of up to 6 percent of global turnover if they fail to verify who’s buying ads (“know-your-advertiser,” or KYA).
- FI will also pilot a real-time ad-registry called FI-ADSPOT with Google, Meta, and TikTok in Q1 2026. Every investment ad targeting Swedish users must be pre-registered and cryptographically signed; anything unregistered will be automatically geo-blocked.
- A Riksdag inquiry (Fi 2025/12) proposes making it a criminal offense to knowingly distribute AI-generated financial endorsements, with penalties of up to four years in prison.
Staying safe: the 2025 investor checklist
- Verify the celebrity. FI posts all known AI-scam ads within 24 hours at fi.se/ai-scam-alerts.
- Check the firm. If the domain is less than two years old or the legal entity is registered outside the EEA, assume it’s high-risk.
- Test the exit. Before investing, try to withdraw €10. Fake platforms almost never allow even small withdrawals.
- Use the “cooling-off” button. Starting 15 November, Swedish banks must offer a 48-hour delay on first transfers to unregulated investment accounts. Use it—it could save your savings.
Bottom line
A half-billion-krona loss is not a rounding error. It’s equivalent to the entire 2024 marketing budget of Sweden’s largest bank. AI hasn’t just automated advertising—it’s industrialised affinity fraud, exploiting trust at scale.
Until platform liability catches up, scepticism is the only real defence. If an investment tip appears out of nowhere, even one delivered by a face you’ve seen on national TV, treat it as guilty until proven innocent.