Instagram's dark funnel: How 50 posts led to a CSAM marketplace

“Is a 14-year-old girl considered a child? Please comment.”

At first glance, it reads like a question. But the Instagram user behind the post does not appear to be seeking an opinion. It is being used to draw attention to content involving the sexualisation and exploitation of children.

And then comes the algorithm. A cliche of the digital age, perhaps, but here it describes something far more consequential: the way users can be drawn deeper into a network of similar content. Like China’s Ruoshui River - said to be impossible to float upon - once you enter, the pull is downward.

Although keyword searches for CSAM-related material remained blocked on Instagram, its recommendation algorithm exposed a significant vulnerability: once users interacted with posts of a similar nature, the platform continued recommending more such content, creating a pathway for repeated exposure to potentially harmful material.

One interaction can lead to another post, another account, and another recommendation.

India Today’s Open Source Intelligence (OSINT) examination of this trail reviewed hundreds of posts and accounts, selecting 50 URLs for detailed analysis. Using established child-safety classifications, 18 were identified as Indecent Images of Children (IIOC), 17 as Child Sexual Abuse Imagery (CSAI), 12 as Child Sexual Exploitation Imagery (CSEI), and three as AI-generated Child Sexual Abuse Material (AIG-CSAM).

This classification was done by India Today based on what was visibly depicted: CSAI, where the content showed elements of sexual activity involving children; CSEI, where children were depicted in sexual exploitation involving adults; IIOC, where a child appeared in an indecent or sexualised pose without sexual activity; and AIG-CSAM, where the material was AI-generated.

The most disturbing finding was not simply that such material existed. It was that some of it remained discoverable.

Instagram restricts searches for certain explicit keywords. Yet after a user encounters several suggestive Reels, its recommendation system can begin suggesting similar content - even when related searches are blocked. Content restricted in search can therefore remain reachable through algorithmic recommendations.

In one case, an Instagram Reel used as a lure showed a young girl being subjected to forced sexual activity, with at least three adult figures around her. The thumbnail carried “Link in bio” and “India Naya Case Part 1”, directing viewers towards an external link. The disturbing thumbnail was being used to attract attention and move users away from Instagram.

In another case, an Instagram profile displayed a Telegram link. The linked channel advertised several categories of explicit videos and claimed to offer 10,000 videos, directing users to contact a seller to choose a package. In yet another example, an Instagram account used a Story to direct viewers to a Telegram group where material was advertised for sale.

The pattern is clear. A deceptive Instagram account posts a misleading or suggestive thumbnail; the viewer is exposed to explicit material; a Story tells them to check a link; Telegram becomes the next destination; private contact follows; samples are offered; and payment is demanded through a QR code, UPI, or another digital method.

Instagram provides the audience. Telegram provides the marketplace. Digital payments complete the transaction.

India Today contacted one Telegram seller during the exercise who claimed to possess “8,400 videos” and said the material would be delivered through a Telegram group. Samples were offered.

For clarity, no samples were downloaded, no material was purchased, and no payment was made to any of the sellers during the course of the exercise.

The scrutiny of Meta is now intensifying

The problem, however, does not appear to be limited to Instagram. India Today also found at least two WhatsApp groups where users were sharing material linked to CSAM, suggesting that such content can circulate across multiple platforms and closed messaging spaces.

That broader ecosystem is also reflected in a new report by the Washington-based Tech Transparency Project (TTP), which found that “Meta ran hundreds of ads containing AI-generated child sexual abuse material (CSAM) on Facebook and Instagram.” TTP said it “found 332 adverts with CSAM that ran on Facebook and Instagram”, of which “84 ran in India”, including 78 that appeared “after the Indian government ordered Meta to remove CSAM-related ads” and content from Instagram.

The report said, “almost all the ads we found start with a photo of a child, and, as the video plays, the photo is manipulated by AI to show them engaging in a graphic sexual act.” Michelle Kuppersmith, executive director of Campaign for Accountability, which runs TTP, said: “If Meta’s ad approval process can’t detect and reject such obviously horrific content before it is shown to Facebook and Instagram users, then Meta’s whole ad review system should be called into question.”

The National Commission for Protection of Child Rights (NCPCR) summoned Meta India’s MD and senior executives after taking “suo motu cognisance” of a BBC Eye report alleging advertisements linked to child sexual exploitation and abuse material on Instagram. The commission issued notice to Meta on July 3, received its response a week later, and then decided to institute an inquiry “to determine the facts of the matter.” On September 9, Meta India MD Arun Srinivas and other senior Meta executives appeared before the commission; the details of the hearing have not been made public.

Additionally, the National Human Rights Commission (NHRC) has sought responses from the Ministry of Electronics and Information Technology (MeitY), the Ministry of Information and Broadcasting (MIB) and the Delhi Police over allegations involving “child sexual abuse material-related ads on Instagram”, though it has not directly issued a notice to Meta.

Meta says, in its July 7 report “Our Work to Fight Child Exploitation on Our Apps,” that it has “zero tolerance” for child exploitation. Following the Indian allegations, it said it removed “additional ads,” disabled accounts and blocked URLs. Meta says it removed “more than 36 million pieces” of child-exploitation content globally last year, while detection of “suspicious off-platform links” and other signals led to 160,000 accounts being removed in India over six months.

The legal obligations are already clear. In December last year, Minister of State for Women and Child Development Savitri Thakur told the Lok Sabha that the IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 impose “due diligence obligations” on intermediaries, including social media platforms. These include that “the intermediary shall make reasonable efforts by itself to cause the users not to host, display, upload, modify, publish, transmit, store, update or share” information that is “harmful to child” or “violates any law.”

In the US, a New Mexico jury imposed $375 million in civil penalties against Meta over broader child-safety allegations. According to the New Mexico Department of Justice, the subsequent court judgment ordered Meta to pay an additional $567 million. The department said, “Combined with the $375 million civil penalty a jury already imposed against Meta earlier this year, today’s ruling brings the company’s total financial exposure in New Mexico to $942 million.” It added that the judgment imposed “sweeping, court-supervised reforms to Facebook and Instagram that will remain in effect for five years.”

Yet the central question remains.

If Instagram can become the place where the audience is found, Telegram the place where the marketplace operates, and algorithms the bridge connecting the two - how much of this ecosystem is actually being stopped before another child becomes part of it?

At the centre of that chain are images and videos involving children.

Beyond the Instagram trail

According to research by California-based child safety organisation Thorn, the scale of the problem is staggering: data from NCMEC shows that registered electronic service providers reported more than 104 million files related to CSAM in 2023. Thorn also found that “2 in 3” victims are abused by someone known to them in their offline communities.

AI adds another layer. The Internet Watch Foundation says the quality of AI-generated child sexual abuse videos has increased sharply, while its 2025 observations describe photorealistic material and warn that “deepfake” child sexual abuse videos have become effortless to generate.