A Global AI Alliance to Stop Child Abuse and Trafficking: HEROES

Children now live online as much as off. Unfortunately predators know this as well. Grooming, sextortion, child abuse and trafficking have shifted to encrypted chats, gaming platforms, and short-form video. The volume is brutal – and still climbing.

In 2024, the U.S. National Center for Missing & Exploited Children (NCMEC) logged 20.5 million CyberTipline reports on child abuse and saw a 1,325% year-over-year jump in tips involving generative-AI technology.

Hotlines in the INHOPE network processed 2.5 million suspected records of child abuse in 2024 alone (which is up by 218% vs. 2023). Europol, meanwhile, coordinated a first-of-its-kind global strike on AI-generated child-abuse material in February 2025, resulting in 25 arrests across 19 countries.

Against this backdrop, an EU-funded consortium called HEROES (2021–2024) pulled together 24 partners from 17 countries – psychologists, software engineers, digital forensics teams, NGOs, and police – to build practical tools and victim-centered protocols that can be deployed now to fight child abuse.

In this article I will go deeper into what HEROES has achieved. Plus, I will also focus on the huge role schools, parents and youth services can play in preventing and reporting child abuse.

Child Abuse Reports Data

The “NCMEC CyberTipline 2023–2024” snapshot shows show startling info on child abuse as you can see below.

Metric	2023	2024	Notes
Total reports received	36,200,000	20,500,000	2024 decline linked to report “bundling” and platform changes.
Adjusted incidents (post-bundling)	—	29,200,000	2024 incidents after de-duplication of viral events.
Files included in reports	—	62,992,859	33,130,449 videos; 28,004,236 images; 1,858,174 other.
Online enticement reports	—	≥546,000	+192% YoY vs 2023.
Child sex trafficking reports	—	26,823	+55% YoY vs 2023.
Reports involving generative-AI	4,700	67,000	+1,325% YoY.
Public reports	—	164,497	Separate from ESP reports.
Reports resolving outside U.S.	—	84%	Illustrates global nature of cases.
Urgent/imminent danger escalations	—	>51,000	Flagged by NCMEC to law enforcement.

Also the data provided by the INHOPE Network Workload and Trends on child abuse in 2024 are more than worrisome.

Metric (INHOPE)	2024
Suspected CSAM records processed	2,497,438
Share with illegal content confirmed	65%
Share that was new content	37%
Common hosting locations	Image hosts; forums
Network size (spring 2024)	54 hotlines across 50 countries

Inside the HEROES Toolkit

Instead of setting up HEROES as yet another lab exercise, the team started by asking investigators and NGOs what would actually help to battle child abuse, then shipped prototypes and guidance. Here’s a list of tools developed, though not all are available yet:

• AGApp (Anti-Grooming App): A mobile application that detects grooming patterns and harmful media in messaging apps and can notify guardians and streamline police reporting. Designed for early, device-side intervention. Not available at the moment
• INDOOR: An AI detector for fake job offers – a common entry point for labor trafficking – built to flag recruitment scams before they hook victims. Not available at the moment
• Citizen Reporting App: Allows citizens to report child abuse incidents of CSA/E and THB to local police, who evaluate and classify the report for further action. Not available at the moment
• Training Curricula: Training programs tailored to prevent child abuse and respond to CSA/CSE and THB. Offered through ICMEC’s Online Learning Platform, accessible free of charge in multiple languages.
• Online THB and CSA/CSE Prevention Programs: Comprehensive guides to educate key stakeholders on online crimes against children, with engaging, culturally appropriate, and accessible content in diverse formats and languages.
• Training Plan for Health Care Workers: A specialized training program to enhance collaboration in investigations and responses to THB/CSA/CSE, promoting better outcomes for victims of child abuse and successful prosecutions.

HEROES sits inside a broader ecosystem – Europol’s EC3, INTERPOL’s ICSE victim-ID database, and a mesh of national hotlines – that turns local leads into cross-border cases.

“How AI Helps” – From Hashes to Classifiers

Child abuse investigators drown in volume. Automation cuts hours off triage and allows scarce victim-ID specialists to focus where impact is highest. INTERPOL’s ICSE database exists for exactly this – comparing imagery globally so a child can be identified and safeguarded faster.

Two detection pillars now operate in parallel.

1. Hash-matching finds known content: Perceptual hashing (e.g., Microsoft PhotoDNA) converts a child abuse image into a signature that will match even if the file was resized or slightly altered. For video, partners use approaches like CSAI Match (YouTube/Google). These systems remove re-uploads fast and generate the majority of platform reports to hotlines.
2. Classifiers surface never-before-seen material: When a child abuse file has no known hash, machine-learning classifiers estimate risk and escalate to trained reviewers. This is how platforms increasingly spot fresh abuse and feed new hashes back into shared sets.

“First Contact Changes Everything” – A Victim-Centered Response

Technology alone doesn’t heal trauma. The first conversation matters a lot. The HEROES manuals and training packages translate 3 principles into day-one practice for local teams:

• Trauma-aware first contact. Clear do’s/don’ts for police and care teams – avoid blame, secure safety, stabilize emotions, and gather the minimum necessary facts with child-friendly procedures.
• Barnahus model. Many European jurisdictions now adapt Barnahus (“Children’s House”) – a single, child-centered location where interviews, medical exams, and victim services happen under one roof, reducing repeated testimony and stress for victims of child abuse.
• Court-ready with less harm. Audiovisual recording of child interviews and specialist interviewers can reduce retraumatization and still preserve evidence quality.

New Threats You Need to Anticipate

Offenders of child abuse iterate amazingly fast – rapidly embracing generative AI, industrialized sextortion, and new lures. This is why the threat model needs to be updated accordingly. Below are 3 new threats.

• Generative-AI abuse. Offenders of child abuse now create synthetic CSAM and deepfake minors at scale. Europol’s 2025 operation showed how quickly this spreads across borders.
• Sextortion factories. Research linked 44 scam compounds in Southeast Asia to hundreds of child sextortion reports and thousands more exploitation signals – industrialized blackmail that blends romance scams, crypto theft, and AI.
• Online enticement spikes. NCMEC reports a 192% rise in online-enticement reporting in 2024, driven by a mix of platform growth and better reporting obligations.

In short, the adversary professionalized. Our response must too.

Europe’s Content-Scanning Debate – Chat Control

The European Commission’s 2022 proposal to prevent and combat online child abuse would allow compelled detection orders for providers. One of the most controversial aspects is chat control, which is the potential use of client-side scanning, where messages are analyzed on a user’s device before they are encrypted, effectively bypassing the core protections of end-to-end encryption.

Supporters argue it’s needed to close gaps; critics warn it breaks privacy and undermines end-to-end encryption, with high false-positive risks. A final vote has been slated this autumn 2025, amid active public campaigns.

Which Stakeholders are Involved?

The fight against child abuse involves many stakeholders. Below is an overview of each group, outlining their core focus and the actions they need to take.

Stakeholder	Core focus	What to do now
Law enforcement (LEAs, digital forensics)	Rapid victim ID, cross-border casework	Integrate hash + ML triage; connect to INTERPOL ICSE; track T3 and rescue rate.
Prosecutors & judiciary	Admissible, child-friendly evidence	Mandate Barnahus-style procedures; record child interviews; standardise chain of custody.
Child-protection NGOs & hotlines (ICMEC, INHOPE members)	Early detection, safe referrals	Use first-contact scripts; sign MoUs with LEAs; route cases to services within 24 hours.
Clinical responders (psychologists, social workers)	Stabilise victims, reduce harm	Apply trauma-informed protocols; use single referral pathway; log outcomes to quality dashboards.
Schools & youth services	Prevention and fast escalation	Teach grooming/sextortion patterns; publish hotline/reporting links; install vetted safety apps.
Transport, hospitality, platform moderators	Spot trafficking signals in the wild	Roll out micro-training; enable one-click hotline reporting; measure referral yield.
Online platforms, cloud & hosting	Fast removal, compliance	Enforce PhotoDNA/CSAI; join trusted hash exchanges; publish error metrics; support reviewer wellbeing.
AI & cybersecurity vendors, researchers	Safer detection tools	Ship investigator-first features; report precision/recall by class; share eval datasets with LEAs.
Parents & caregivers	Home and device safety	Set device agreements; enable alerts; rehearse “pause—don’t pay—report.”
Journalists & advocacy groups	Public awareness, policy scrutiny	Use verified stats and timelines; link training/hotlines; spotlight due-process safeguards.
Funders & programme officers (EU, foundations)	Scale what works	Tie grants to KPIs (T3, rescue rate, FPR); fund cross-border taskforces; audit quarterly.
Data Protection Authorities (DPAs)	Rights, legality, oversight	Define lawful bases; require minimisation and audit trails; review false-positive rates.
Policy makers & standards bodies (EU CSA file, AI Act, DSA)	Legal guardrails and transparency	Clarify detection orders; protect E2EE with safeguards; mandate reporting on errors and appeals.
Europol EC3 & INTERPOL (ICSE)	International coordination	Schedule joint ops; expand ICSE connectivity; exchange leads securely with audit logs.
ISPs & telecoms	Network-level support	Enable rapid data preservation; maintain lawful disclosure channels; track turnaround times.

Checklists for Schools, Parents and Youth Services to Help Fight Child Abuse

For schools, youth services, and parents there exist 3 golden rules to apply in the fight against child abuse.

1. Teach the tactics: Explain grooming scripts, sextortion patterns, and fake job red flags; rehearse “pause-don’t-pay-report” reactions.
2. Instrument devices: Where lawful, deploy on-device safety apps (e.g., AGApp-style functionality) with transparent family agreements and escalation paths.
3. Map help fast: Post local hotline numbers, reporting portals, and Barnahus/child-services contacts on every school device and LMS.

In the below checklists I outline further for each what the specific steps should be in the battle against child abuse.

Schools: act in class, on devices, and at the gate

Do this in 7 days

• Publish a one-page Safeguarding & Reporting Ladder on the school site and LMS. Place the same sheet in classrooms and staff rooms.
• Run a 45-minute Digital Harm Drill for grades 10–14 (ages vary by country): what sextortion looks like, how grooming starts, how to capture evidence, where to report.
• Appoint a Single Point of Contact (SPOC) for abuse disclosures. Add their name and direct line to student planners.
• Turn on enforced filtering on school networks. Block anonymous proxies and high-risk file sharing. Log incidents to a restricted inbox.
• Add platform reporting links (e.g., “Report” on social/video apps when you see signs of child abuse) to the LMS homepage.

Build in 30 days

• Implement Barnahus-aligned interviewing (one trained interviewer, recorded once, child-friendly room).
• Integrate dual-rail triage in IT: hash matching for known illegal content; classifier escalation to the SPOC for unknown files (human review required).
• Launch micro-modules for all staff: 10 minutes each on grooming scripts, sextortion playbooks, fake job lures, and first-contact do’s/don’ts.
• Establish MoUs with the local child-protection service, police, and the national INHOPE hotline.

Keep doing each term

• Run parent evenings on sextortion and deepfakes. Show real scam flows (redacted).
• Audit “Time-to-Triage (T3)” from detection to first action. Target under 60 minutes during school hours.
• Stress-test your incident plan. Simulate a disclosure during lunch, a device seizure, and an urgent referral.

Classroom red flags

• Sudden secrecy around a new “online friend.”
• Requests to move chats off-platform.
• Requests to go to a theme park, cinema or other place.
• Pressure to send images, then threats to expose.
• Gifts, crypto, or game credits sent to the student.
• Talk of “models”, “brand ambassador”, “easy jobs.”

First-contact script (teacher/SPOC)

“You did the right thing by telling me. You are safe here. You are not in trouble. I will only ask what I need to keep you safe. We can pause anytime. We will not make you repeat this to many people.”

Reporting ladder (EU context—adapt locally)

1. Secure the child and device. Do not forward files.
2. Notify the SPOC and child-protection lead.
3. Report online content to your national INHOPE hotline; report missing/exploitation emergencies via 116 000 (Missing Children Europe).
4. Offer the child access to a counsellor; inform caregivers according to policy.
5. Preserve logs, timestamps, and URLs for law enforcement.

Parents & caregivers: set rules, spot lures, respond fast

Set the ground rules tonight

• Create a device agreement: where devices live at night, who approves new apps, what happens when something feels wrong. Both sides sign.
• Turn on platform-level reporting and content filters. Update devices. Disable unneeded direct messages for younger teens.
• Agree on a one-tap plan: when pressured or threatened—stop, screenshot, block, report, tell an adult. No paying ransoms. No negotiating.

Talk tracks that work

• “Scammers use shame. We remove the shame by talking. I promise support first, solutions second.”
• “No one loses privileges for reporting a problem. Silence helps offenders; talking stops them.”
• “A real friend never rushes you to send images or move to a secret app.”

Home red flags

• Mood swings after messages ping.
• Money, vouchers, or crypto entering or leaving accounts.
• New profiles that look too perfect or ask to hide the relationship.
• Sudden fear of someone exposing images.

When sextortion hits

1. Stop responding. Do not send more images or money.
2. Save evidence: usernames, URLs, payment handles, time, screenshots.
3. Report on-platform and to your national hotline; contact police if threats escalate.
4. Tell the school SPOC if peers are involved.
5. Sit with your child. Remove blame. Arrange follow-up counselling.

8 Parent Device Agreement

1. We keep devices out of bedrooms at night.
2. We approve new apps together.
3. We never share private images.
4. If someone pressures or threatens, we stop and tell an adult.
5. We save evidence.
6. We report on the app and to our national hotline.
7. We never pay ransoms.
8. We face problems together. Always.

Youth services: catch signals of child abuse outside school

Where you see it
Sports clubs, libraries, after-school programs, transport hubs, shelters, and community centers.

Equip your team

• Train front-desk and volunteers to spot fake job lures and grooming behaviors.
• Post the Reporting Ladder at staff-only desks.
• Provide a private space for disclosures. Keep water and a simple “You’re safe here” card ready.

Fast protocol

• Stabilise first. Move to a quiet room. Use the first-contact script.
• Capture only essential facts. Time, place, platform, type of pressure or threat.
• Trigger your referral: child-protection service, national hotline, then police when urgent.
• Do not promise secrecy; promise safety and support.

Partnerships to lock in

• MoUs with the nearest Barnahus/Children’s House, local police, and the national hotline.
• A 24/7 escalation phone chain for after-hours incidents.

Digital safeguards everyone can apply

• Reduce surface area: close unused DMs, hide contact lists from non-friends, restrict duets/stitches if using short-video apps.
• Limit discoverability: remove phone/email from profile searchability. Use private accounts for minors.
• Control payments: block in-app purchases and non-family transfers on youth devices.
• Keep evidence intact: never forward illegal content; store originals for law enforcement only.
• Teach the scam flow: friendly approach → flattery → nudge to share → threat → demand for more or money.

Protect Children, Equip Adults, Move Faster

Child abuse and trafficking thrive on speed, secrecy, and scale. Your response must outpace all three. This article mapped the essentials: dual-rail detection for known and new content, trauma-aware first contact, cross-border casework, and clean governance. Together, these turn scattered effort into a coordinated shield that actually stops harm.

AI isn’t the hero. People are. The tools – hash matching, risk classifiers, device triage – only matter when investigators, clinicians, and hotlines work as one. HEROES showed the model: ask practitioners what they need, build to fit real workflows, and measure outcomes that matter – time to triage, rescue rate, false-positive rates, and referral speed.

Policy debates will continue., however children cannot wait for perfect consensus.

Now, schools. You sit on the earliest signals. You see new “friends,” sudden secrecy, and the panic that follows threats to expose images. Turn that proximity into protection. Name a safeguarding lead, post a one-page reporting ladder everywhere, and rehearse first contact until it sounds natural under pressure. Wire IT for dual-rail detection. Prepare a child-friendly interview room so a child tells their story once.

Bring families with you. Share a simple device agreement and a one-tap plan: stop, screenshot, block, report, tell an adult. Host short parent sessions that show how sextortion actually unfolds. Remove blame throughout. Promise support first, solutions second.

Finally, commit in writing. “We respond within 60 minutes. We record once in a child-friendly room. We never forward illegal content. We preserve evidence securely. We brief every staff member every term. We inform parents within policy. We review outcomes monthly and improve them.” Put that pledge in the staff room and on the LMS. Live it.

This is how you win ground back: teach the tactics, detect early, escalate cleanly, and support without judgment. When schools lead, children stay safer, cases move faster, and offenders lose room to operate. Start today. Keep going tomorrow.

---

---