Tests: Meta's recommendation systems promote underage sex content on Instagram and Facebook Groups, despite Meta setting up a child-safety task force in June
The report follows a June investigation that found Instagram connecting accounts devoted to underage sexual content and Meta's subsequent pledge to increase controls through a child-safety task force. The new tests put pressure on whether those controls changed distribution rather than merely account-level enforcement.
Later coverage makes this a persistence story: [[a:867785|subsequent tests found Instagram still recommending sexual videos to accounts listed as age 13]], while reporting also raised concerns about child exploitation through Meta's subscription tools.
First-order effects
Meta's safety task force and recommendation controls face a direct credibility test, as the reported recommendations span both Instagram and Facebook Groups.
Accounts interacting with this material can be routed toward more underage sexual content by the platforms' discovery systems, according to the tests.
Second-order effects
Meta must weigh tighter recommendation and group-discovery safeguards against the engagement objectives that ranking systems are designed to serve.
If repeated testing continues to find harmful recommendation pathways after safety commitments, platform accountability will increasingly turn on measurable ranking-system outcomes rather than stated policies.
The case supports a broader shift toward age-sensitive product testing and access-control expectations for social platforms, though the eventual regulatory response remains uncertain.
The trend: Child-safety scrutiny is moving from whether platforms prohibit harmful content to whether their recommendation systems still amplify and connect it.
Meta is *also* simultaneously trying to have the FTC declared unconstitutional, echoing far-right arguments being made in other regulatory/ administrative law cases.
Friendly reminder: our EdTech Safety Scores (found in https://appmicroscope.org/) call out Meta as substantial risk to users. When you see it in the label (under “aggregator platform"), it means we observed traffic from the app to the Meta's servers. (you may also see it in SDKs.…
Meta's recommendation systems just can't stop promoting child sexual abuse at scale. The story today from @katherineblunt and I backs that statement up. But exploring the scope of the problem a little more on here seemed worthwhile.
This time, researchers with @stanfordio found that Meta was suggesting variations on pedophile hashtags to get around its own safeguards. Under current interpretations of Section 230, Meta cannot be held accountable.
Yet another piece from @JeffHorwitz and @KatherineBlunt reveals that Meta is STILL helping pedophile networks share material on Facebook and Instagram. If Meta can't build safe algorithms, it shouldn't be using them to shape its platforms. https://www.wsj.com/...
Despite launching a Child Safety Task Force in June, the company hasn't fixed the problems underlying the issuies we identified, such the pedophilic “accounts you should follow” recommendations to autocomplete suggestions that users look for “child pornography links” in search.
Meta has described the pedophilic recommendations problem we identified as unrepresentative. But on a platform with three billion users, a fraction of a percent adds up to millions. And its what the members of Facebook's massive pedophile Groups have been consuming every day.
Parents: I think it's important for all of you to know this is happening because the FTC is working to limit surveillance capitalism on your kids by Facebook and Instagram. And they're doing it because owner Meta has repeated abused privacy leading to billions in settlements. [im…
Meta is having trouble keeping pedophiles off Facebook and Instagram. Despite stepped up enforcement, its algorithms keep promoting problem content. https://www.wsj.com/...