Meta is liable for the Instagram scam ads it chooses to show

A familiar face recommends an investment on Instagram or Facebook. The link opens a WhatsApp group, and a supposedly safe tip is waiting there. The money is gone afterwards. For months, fake profiles used the name and photos of Thomas Kehl, the face of the German personal finance platform Finanzfluss, for that scheme. On Sept. 16 the Regional Court of Frankfurt am Main ordered Meta to stop the practice, to disclose information and to pay damages (case 2-06 O 234/25).

The plaintiffs had used every tool Meta offers. They filed reference images in the Brand Rights Protection Tool, reported roughly 256 violations in a single month and assigned a full-time employee to monitoring. Takedowns still took 14 days, 20 days and in one case 62 days. New fakes appeared almost daily after each one, some of them a deepfake video. Investors who had lost money then turned to Finanzfluss for help.

The ad auction costs Meta its hosting exemption

The reasoning matters far beyond this case. Until now Meta counted as a neutral hosting provider and could rely on Article 6 of the Digital Services Act. The company was liable only if it failed to act after a report. The Frankfurt chamber disagrees. According to the court, Meta uses an advertising auction with criteria it sets itself to determine which ad reaches which user, and its ranking algorithm does the same for ordinary posts. A service that decides how content spreads is no longer a passive host.

The court states this directly. Meta is the party that determines the conditions under which content is distributed or withheld. It relies on a ruling by the Court of Justice of the European Union from June, which found that algorithmic distribution can remove a platform's hosting exemption. Meta is therefore liable as a perpetrator rather than an intermediary, which also allows damages and monetary compensation for Kehl. Each further violation can cost up to 250,000 euros, and the chamber set the amount in dispute at 300,000 euros.

Chronological feeds are not covered

The court explicitly limits its finding to services that rank content. Providers that leave the selection to their users do not exercise that kind of control. The judgment names classic online forums as well as purely chronological feeds such as those on Mastodon and Bluesky. A provider therefore decides for itself whether to sort by algorithm and take on responsibility for what other people post.

People who fell for such an ad get nothing out of it

The plaintiffs here are the people whose face was stolen, not the investors who trusted the ad. Anyone who lost money in one of those WhatsApp groups cannot derive a claim against Meta from the Frankfurt judgment. They still have to pursue the fraudsters personally, and those are usually based abroad. The decision still matters, because the injunction also covers essentially identical variants, including profile names with added digits or special characters. Not every new copy has to be litigated separately.

Until the filtering obligation takes effect, spotting the scam yourself is the only protection. Investment tips from celebrities in social media ads are almost always fake, and a redirect into a WhatsApp or Telegram group is the clearest warning sign. Sender details and layout can look entirely convincing, as a recent Trezor phishing wave showed, where the emails came from the genuine sending address. Meta does run its own AI against fraudulent messages, but Scam Alert in WhatsApp checks only senders outside your contact list. Apple at least builds a bank fraud defense into iOS 27, although it ships switched off.

The judgment is not final. Meta can appeal to the Higher Regional Court of Frankfurt, and the question of algorithmic control will probably end up before Germany's Federal Court of Justice.