SAFE Platforms Act: a good start (but there’s a big loophole)

previously: fix liability and ad libraries: anti-fraud policy ideas from Canada

, intended to address the Big Tech scam ads problem.

A good start—especially the part about unambiguously identifying an advertiser to the people who get their ad—but because of how the ad targeting ML works, there is one big loophole.

Platforms shall provide users with access to basic information about advertisers whose content they view

Emphasis added.

The problem here is that ad personalization ML “learns” to target ads not to the people who might report a scam, such as regulators or the holders of a trademark that the ad infringes, but to people who might fall for the scam.

So a transparency requirement that only affects ads that the ML already decided to show you is inadequate.

In general, transparency must apply to all ads on the platform, not just the ads pre-selected as those a particular individual is likely to click.

This bill needs an ad library section. Right now, the platform ad libraries, such as Google Ads Transparency Center and Meta Ad Library, are designed with some fraud-friendly features that make it easier for deceptive sellers to hide. Some technically straightforward but high-impact fixes include:

• Ad libraries should not block crawling by independent image search sites and trademark monitoring companies.
• An ad should not serve to regular people until after it had been available in the platform ad library for a time period to be set by regulations.
• All substantial ad variations should be available.
• Index all text in the ad (using OCR if needed), not just “search by advertiser or website name.”
• Show, and allow searching by, intermediary companies involved in placing an ad, not just the advertiser.
• For each ad creative, link to all accounts running the same or substantially similar ads.
• Cross-link accounts that are using the same domains in landing pages.

Legit advertisers have several incentives to check platform ad libraries. The biggest is to protect their customers from fakes, especially fakes that are bad enough not to work, but good enough to make customers think they’re real and try to get support from a real manufacturer or retailer.

Legit advertisers can also keep their own ad costs down by getting the scams kicked off. Scammers act as shill bidders to drive up platform ad revenue, so platforms are disincentivized to remove them without outside attention.

Other public interest organizations also have an interest in protecting people from deceptively sold goods. Platform ad libraries must be designed to facilitate that help, not make it harder.

The Canadian Anti-Monopoly Project recommends an archival requirement for platform ads, which is a good start. But IMHO it’s important to consider the information needs of would-be scam fighters and the scam protection features of platform ad libraries when legislating in this area.

Bonus links

German court rules Meta liable for scam ads on Facebook and Instagram by Alina Maria Stan. Finanzfluss, a German personal finance platform, and its co-founder Thomas Kehl filed the lawsuit. Scammers used the company’s trademarked logo and Kehl’s photo without permission in posts promoting investments the court called allegedly fraudulent… Meta argued it should not be responsible for content it did not know about, which is a defense allowed for hosting services under the EU’s Digital Services Act.

This is how to search now that Google Search is gone: a guide for family and friends by Kim Crawley. (Lists some more problems with Google Search besides just the crime ads thing.)

Approaching A 10 Second Linux Kernel Build by Michael Larabel. (The new PCs that the AI oligopoly won’t let us have are pretty sweet. And this is with Rust in the kernel, too.)