Meta's controversy over the circulation of paid advertisements depicting child sexual abuse material (CSAM) on its Instagram platform made headlines earlier this week, drawing attention to how such harmful content slips through gaps in regulation, legislation and platform enforcement.
NDTV Profit spoke to cybersecurity experts Kunal Bhogal, chief operating officer at IIRIS Consulting, and Anurag Singh, chief executive officer of RAH Infotech, to understand the emerging technologies cybercriminals use to evade detection while circulating CSAM.
Experts described AI as a "game changer" in both the creation and distribution of such content. According to the Internet Watch Foundation, AI-generated deepfake CSAM surged 400% within six months in 2025. Organised criminal networks increasingly operate as decentralised groups spanning multiple countries, using encrypted messaging platforms, cryptocurrency payments and dark web hosting.
Getting Past Detection Systems
Cybercriminals use a range of techniques to evade detection, including coded language, altered images, indirect links, cross-platform movement and automated systems that create fake accounts, distribute links and redirect users from open platforms to encrypted or private groups.
These tactics often exploit the limitations of AI-powered moderation systems, which remain effective at processing enormous volumes of content but can struggle to understand context or identify newly created material. Experts noted that technology companies rely heavily on AI because of the sheer scale of content uploaded every day, while also reducing the psychological burden on human moderators exposed to disturbing material.
ALSO READ: What Is WhatsApp's Meta Business Agent? Here's How The New AI Tool Automates Customer Chats
AI systems can perform first-level detection at scale by identifying known content through digital hashes, detecting suspicious language patterns, analysing account behaviour and tracking repeat offender networks. Cybercriminals, however, continually test these systems to identify blind spots and exploit their weaknesses.
"AI is brilliant at spotting material that's already been identified because every known image has a digital fingerprint. But show it something new, or disguise it with coded slang, and it stumbles," said Kunal Bhogal, COO of IIRIS Consulting.
AI can also miss contextual cues, allowing criminals to exploit moderation systems. They may use stolen payment cards, fake advertiser accounts or submit seemingly harmless advertisements before later changing the destination or linking users to prohibited content.
According to Anurag Singh, CEO of RAH Infotech, CSAM networks frequently rely on coded language, manipulated images, indirect links and movement across multiple platforms to avoid detection.
What Companies Can Do To Improve Moderation
Experts said advertising platforms should have visibility into the advertiser, payment method, campaign creative, landing page and targeting patterns, creating multiple opportunities to identify suspicious activity.
"Where such content is found to have slipped through, it usually points to gaps in one or more areas: advertiser verification, automated review, landing-page checks, detection of coded language, or the ability to connect repeat signals across accounts," Singh said.
In a recent public statement, Meta said it uses both automated and manual review systems to enforce its advertising policies. The company also said it monitors advertiser behaviour alongside individual advertisements.
"Beyond reviewing individual ads, we also monitor and investigate advertiser behaviour, and may restrict advertiser accounts that don't follow our advertising standards, community standards or other Meta policies and terms," the company said.
Meta also said it uses advanced AI to proactively detect policy-violating content and individuals. However, experts recommend a layered moderation approach combining AI with human oversight.
ALSO READ: Meta Says Will Strengthen Review Process Amid Reports Of Child Sexual Abuse Ads On Instagram
"The better model is a layered one: AI for scale, human review for sensitive decisions, specialised child-safety teams for escalation, and faster reporting pipelines to law enforcement. In cases involving paid ads, the threshold for review should be even higher because money, targeting and platform approval are involved," Singh said.
Experts also warned that some companies reduce human moderation to cut costs, creating gaps that automated systems alone cannot adequately address.
"Companies know the human layer matters and still skimp on it because it's expensive," Bhogal said.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
