More than half of social media child safety features aren’t working as advertised, new research finds

Researchers tested safety tools meant to protect young people across four major social media platforms, and say that more than half of them failed to perform as advertised. (skynesher/E+/Getty Images via CNN Newsource)

NEW YORK — Social media giants have for years touted their growing slate of safety tools and protections as proof that they prioritize young users’ wellbeing.

But more than half of those protections don’t work as advertised, new research finds.

Researchers at the Cybersafety Research Center tested 86 youth safety features across TikTok, Instagram, Snap and YouTube and examined whether they worked as described and whether children could realistically find and use them. Only 35 of those features — just over 40 per cent — successfully met both criteria.

The companies largely disputed the report’s findings, arguing that their features work as intended or that the tests did not represent typical use of the platforms by kids and teens.

YouTube and Instagram parent company Meta were also found liable this year of intentionally addicting and harming young people. All four companies face thousands of lawsuits with similar claims, which they have disputed.

“If you are a parent, you should know that we have found systemic issues with the design and implementation of many of these features,” the report, published Monday, states. To the Cybersafety Research Center, a joint initiative from New York University and Northeastern University, the findings show social media harms to kids “are not hypothetical, and that when they do occur the consequences can be irreversible.”

The findings call into question social media companies’ claims that heavy investments in new tools and features have made their platforms safe for young people. And they come amid a renewed push for more federal regulation of social media, with executives from some of the top firms expected to be called to testify on Capitol Hill again this summer.

The tests

Researchers created two sets of test accounts: those with birthdates associated with minors to test child safety features, and adult accounts ages 25 and older to test restrictions on interactions with children. They tested whether a young person using the platforms normally would encounter the features, whether a teen attempting to circumvent the features would succeed and whether an adult user could bypass restrictions on messaging minors.

Features were categorized as failures if they were: buried in settings and hard to find, broken because they did not function to prevent harm as advertised, or both. Nine features were also labelled as “missing” because researchers couldn’t trigger them even when they tried. For example, comments between teen accounts on Instagram that included bullying language — like curse words and “no body (sic) likes you” — did not trigger the platform’s advertised prompt for the user to rethink their comment.

Instagram said the “pause to rethink” prompt is not designed to appear if the user who made the post and the user commenting follow each other.

“Safety features, in order to be effective, need to be on by default or easy to activate, be resilient to normal teenager use, and should demonstrably protect against harm,” the report states.

The report did not look at parental control features, reviewing only tools that would apply directly to or can be turned on by kids and teens. The platforms say parental controls also contribute to youth safety and make it harder for kids and teens to bypass time restrictions and other features.

The number of features available to be tested varied by platform, with Instagram having the most at 29. The feature failure rates by platform were: Snapchat 73 per cent, Instagram 66 per cent, YouTube 55 per cent and TikTok 50 per cent.

The features

All four platforms say they block children from searching for dangerous content and instead direct them to support resources. But researchers found they fell short of doing so in practice.

After a TikTok test account registered to a minor searched for material about disordered eating and self-harm, the app’s search function then suggested terms such as “how to pretend to eat your food” and “razor blade skin,” according to the report. As a child test account began typing “eating disorder” into Instagram’s search bar, the app automatically offered alternate search terms with deliberate misspellings that could skirt content restrictions. Misspellings also worked to circumvent restrictions on Snapchat.

A TikTok spokesperson said the app’s teen accounts settings “come with over 50 preset safety features and settings automatically turned on, with additional choices for parents through our easy-to-use Family Pairing tool. Our internal review confirms these features are working as intended, and we welcome the opportunity to help the authors of this report better understand how our app works.”

Researchers rated YouTube’s search content restrictions as a success. But they said YouTube’s efforts to redirect young users to resources failed because a child user could click away from the restricted screen and continue to view content they’d searched for. YouTube says that if a teen dismisses that resource direction screen, they will still only see content rated as safe for their age group.

“We’ve spent over a decade building industry-leading parental controls, which is why 84 per cent of parents who have used YouTube supervised account tools said they agree that these tools give them confidence that their child is accessing a safer and more controlled digital environment,” a YouTube spokesperson said in a statement about the report. “We will continue to strengthen these protections and innovate to protect families who use YouTube.”

Researchers also raised questions about Snapchat and Instagram’s efforts to block adult strangers from messaging with young users. On Snapchat, researchers using an adult test account said they were able to find and message a child account with no restrictions. The child account “received the friend request and upon accepting it was able to view the history of messages that the adult had sent them with no warnings,” the report states.

On Instagram, researchers found adults who a child didn’t follow were unable to initiate messaging conversations with that child, saying the app’s restriction works as described. But they argued the feature is compromised because a child could message an adult they don’t follow and the adult can then respond — with no restrictions or warnings.

The report points to a 2021 press release from Meta which says that Instagram restricts people over 19-years-old from sending private messages to teens who don’t follow them. “We find that this is not strictly true; the adult is able to send messages to a child unrestricted after contact has been initiated, even if the child does not follow them back,” the report states.

Meta told CNN a teen messaging an adult who does not follow them indicates the young person wants to connect.

The researchers say they alerted Instagram and Snapchat to the child messaging concerns and other “critical vulnerabilities” before releasing the report.

A Snapchat spokesperson told CNN the company cares “deeply about the safety, privacy, and well-being of all Snapchatters, and our teams have worked for years to build safeguards, launch safety tutorials, and partner with experts … we are continually evaluating and strengthening our protections.” The spokesperson added that many of the report’s findings were based on “intentionally taking actions to bypass protections that are not representative of the typical user experience.”

Many of the platforms also offer time limits or “take a break” reminders to prevent kids from endlessly scrolling. But on Instagram, YouTube and TikTok, the prompt telling users to take a break includes an option to “snooze” the reminder and return to the feed. Meta and YouTube said the feature works as intended and added that time limits set up via parental controls could not be dismissed.

“This report is fundamentally flawed and demonstrates a basic misunderstanding of how our tools work,” a Meta spokesperson said in a statement. “The authors include vague claims that our features are broken but, in the vast majority of cases, either misrepresent those features or fail to provide any examples or evidence. The reality is that with Teen Accounts, teens are seeing less sensitive content, experiencing less unwanted contact, and spending less time on Instagram at night.”

Meta argued that some of the report’s findings about “missing” features may have simply been because they did not encounter issues that would have triggered the tools. For example, researchers said they did not see any “sensitive content” warning screens on potentially problematic posts; Meta said no sensitive content may have appeared in their feeds.

Some features were successful. For example, if users under the age of 13 try to sign up for a TikTok account, they’re automatically directed into a “TikTok for Younger Users” experience that is view-only and removes potentially dangerous app features, such as search and messaging. That means kids don’t have an opportunity to try to sign up again with a fake, older birthdate on that device.

And minor accounts automatically default to private settings on Instagram, which researchers said, “could encourage users to maintain private accounts without requiring the user to make an informed decision.”

Those successes prove that it is possible to design effective safety youth features, the researchers said. Companies should create safer online platforms by trying to reduce overall risk, they wrote, “rather than filtering a dangerous experience.”

By Clare Duffy, CNN