Tests show CSAM still spreads and gets recommended on Twitter, including known material that's easier to detect; one video had 122K+ views in just over a month
New York Times
Context & Ripple Effects
The Times' tests land two months after Elon Musk's late-November pledge to stamp out child sexual abuse material on Twitter — and find that not only is known, hash-matchable material still live, some of it surfaces through the recommendation algorithm itself, with one video clearing 122,000 views in just over a month.
The finding sits inside a broader pattern the corpus keeps confirming: an earlier study of COVID misinformation showed most flagged content simply stayed up on these platforms, and the Internet Watch Foundation has since reported record and rising webpage counts for CSAM overall.
First-order effects
Twitter's advertising base faces direct brand-safety exposure — with reported ad revenue of $1.08B in Q2 2022 against roughly $4B in full-year 2021, any advertiser exodus over safety failures lands on the company's core income line.
Known-material detection, the easiest category to catch because it matches existing hashes, demonstrably failed at scale — putting Twitter's trust-and-safety operations under immediate scrutiny from child-safety groups and researchers who now have a repeatable test method.
Second-order effects
Competing and adjacent platforms get pulled into the same audit cycle: researchers ran the same kind of test on Mastodon's top instances months later and found hundreds of known or suspected files, showing decentralization without centralized hash-matching infrastructure reproduces rather than solves the problem.
Child-safety organizations such as the Internet Watch Foundation gain leverage to demand mandatory, auditable detection practices, since their own data shows total CSAM webpages climbing year over year regardless of which platform hosts them.
Third-order effects
If algorithmic recommendation of known CSAM becomes a documented pattern rather than a one-off, the likely endpoint is regulation that mandates hash-matching and detection reporting across all major platforms — including federated networks that currently lack a single accountable operator.
The deeper structural risk is that moderation capacity shrinks faster than content volume grows: as platforms cut trust-and-safety staffing while adoption of AI-generated media accelerates, detection tooling lags the material it must catch, leaving enforcement dependent on external watchdogs rather than the platforms themselves.
The trend: Platform moderation is falling behind both content volume and content migration — from centralized feeds to federated instances — making external audits and regulatory mandates the de facto enforcement mechanism for child-safety compliance.
Is Elon Musk lying about child sexual abuse on Twitter? — Musk says that cracking down on child exploitation content is “priority #1” for Twitter. — But the facts say otherwise. — Not only does child exploitation persist on Twitter. …
Elon Musk said that ridding Twitter of child sexual exploitation material was “priority #1.” @mhkeller and I set out to understand the scope of the problem: https://www.nytimes.com/...
Twitter's head of safety Ella Irwin “did not dispute that child sexual abuse content remains openly available on the platform”. Musk said that child exploitation was his #1 priority in November and then sacked most of the Twitter trust and safety team. https://www.nytimes.com/...
Since M*sk pledged to make removing CSAM “priority #1,” the problem seems to have gotten at least somewhat worse. I'd hate to see how priority #2 is going. https://www.nytimes.com/...
“people on dark-web forums discuss how Twitter remains a platform where they can easily find the material while avoiding detection” https://www.nytimes.com/...
In January, Twitter said it suspended 404,000 accounts for CSAM-related violations. NCMEC, the clearinghouse that tech companies are supposed to report these violations to, said it only got about 8,000 reports from Twitter that month. https://www.nytimes.com/...
Because Twitter recently added view counts to tweets, we could see that some of the material, which remained on Twitter's platform for days and sometimes up to a month, attracted hundreds of thousands of views.
@mhkeller We used an automated program written by @mhkeller to scan for content hashed by PhotoDNA, which identifies exploitation material that is known to authorities and has typically been widely circulated online. Within a few hours, we started getting hits.
“The material wasn't difficult to find. In fact, Twitter helped promote it through its recommendation algorithm — a feature that suggests accounts to follow based on user activity.” https://www.nytimes.com/...
New w/ @kateconger: Over 120k views on a video of child sexual assault. A recommendation engine pushing abuse-related content. Unpaid bills for detection software. We looked at Twitter's efforts against child sex abuse since Elon Musk made it “priority #1” https://www.nytimes.com…
“To find the material, Twitter relies on software created by an anti-trafficking organization called Thorn. Twitter has not paid the organization since Mr. Musk took over... Twitter has also stopped working with Thorn to improve the technology.” https://www.nytimes.com/...
Old Twitter was a cesspool of child sexual exploitation and yet the New York Times was silent. Now with @elonmusk at the helm actively attempting to rid the platform of the content the Times is critical of the billionaire - they hate him more than CSE. https://www.nytimes.com/...
NYT built a bot to study CSAM on here. such “material wasn't difficult to find. In fact, Twitter helped promote it through its recommendation algorithm” https://www.nytimes.com/...
Absolutely devastating indictment of Elonza. “they [Twitter] are not doing the basics” “Twitter relies on software created by an anti-trafficking organization called Thorn. Twitter has not paid the organization since Mr. Musk took over” @wokbonds https://www.nytimes.com/...
An NYT review of Twitter found child porn and exploitation “persisted on the platform, including widely circulated material that the authorities consider the easiest to detect and eliminate,” despite Musk's promise to do so https://www.nytimes.com/...
“All the while, people on dark-web forums discuss how Twitter remains a platform where they can easily find the material while avoiding detection, according to transcripts of those forums from an anti-abuse group that monitors them.” https://www.nytimes.com/...
OK, so Elon didn't reduce child porn on Twitter. And yes, he called FBI reimbursements for aiding child porn investigations censorship payments. But he said he'd do it, accounts he pushed into people's feeds said he did, and isn't that what really matters? https://www.nytimes.com…
“But a review by The New York Times found that the imagery, commonly known as child pornography, persisted on the platform, including widely circulated material that the authorities consider the easiest to detect and eliminate.” https://www.nytimes.com/...
“After Mr. Musk took the reins in late October, Twitter largely eliminated or lost staff experienced with the problem and failed to prevent the spread of abusive images previously identified by the authorities...” https://www.nytimes.com/...