The behavioural data behind autonomous protection.
Research that informs the behavioural ontology — grooming patterns, sextortion progression, athlete and influencer abuse, and the linguistic markers that machine systems detect long before a human reviewer would. Sourced from law enforcement, academic studies, and child safety organisations.
The hidden layer of youth bullying
On 9 August 2026 the Australian Government launched a national anti-bullying campaign built on the claim that bullying is harder to spot than parents think. Its own research explains why that is true. Youth bullying has migrated into private messaging, and the settings children name most often are the ones no adult is present in. This is the most consequential and least visible shift in youth safety, and it is the layer that awareness alone cannot reach.
Where a child’s most recent cyberbullying actually happened
eSafety Commissioner. The 696 cyberbullied children within a survey of 3,454 Australians aged 10 to 17, asked about their most recent experience. Fieldwork December 2024 to February 2025.
Children could choose more than one setting, so these shares overlap and do not sum to 100%.
were last cyberbullied in an online group chat, the single most common setting. Only 8% were last targeted on someone else’s social media page.
eSafety Commissioner — Snapshot: Cyberbullying (n=696)of Australian children aged 10 to 17 have been cyberbullied at some point, and 38% within the past twelve months.
eSafety Commissioner — 3,454 children surveyed Dec 2024 to Feb 2025of parents whose child had been treated hurtfully online were aware of it, even though 69% of those children said they had told them.
eSafety Commissioner — Mind the Gap, 2021 fieldwork (3,651 children, 3,590 parents)of children were cyberbullied by someone they knew in real life, most commonly a friend (62%).
eSafety Commissioner — Snapshot: Cyberbullying (n=696)of all Australian children aged 10 to 17 have been told online to hurt or kill themselves, or that they should die. 7% in the past twelve months.
eSafety Commissioner — Snapshot: Cyberbullying (n=3,454)increase in cyberbullying reports to eSafety between 2019 and 2024, from 536 to 2,978. 46% of the 2024 reports concerned children aged 13 and under.
eSafety Commissioner — Safer Internet Day release, Feb 2025The platforms where it happened
Share of children naming each platform for their most recent cyberbullying experience. Messaging and gaming lead, and both are places where conversation is private by default.
The severe end moves faster than any reporting process
Australian Institute of Criminology, Trends and Issues No. 712, published February 2025 from a survey of 1,953 Australians aged 16 to 18 conducted in mid-2024.
of Australian 16 to 18 year olds had experienced sexual extortion in their lifetime. Two thirds of them had only ever met the offender online.
were threatened less than a day after first contact. 58.6% were threatened within a week.
were asked to move the conversation to another platform, most often Snapchat (31.9%), Instagram (29.2%), WhatsApp (16.7%), Facebook and iMessage (9.7% each).
The Institute’s own conclusion is that the use of encrypted communication to carry out sexual extortion introduces challenges for detection and deterrence. Separately, eSafety found that one in eight child sexual abuse material complaints it analysed involved an offender directing a child to perform acts on camera, with offenders described as sliding into direct messages on online games and social media. A response process that begins when an adult finds out is not built for a harm that escalates from first contact to threat inside twenty four hours.
Why this is the subtle end of youth bullying
- ▸The Anti-Bullying Rapid Review found that cyberbullying and online harms are “much harder to escape from and can cause particularly acute suffering through their invasive nature and ability to spread material rapidly and widely”. (Final Report, October 2025)
- ▸The Review cites a figure of 84% of students bullied online also being bullied in person, so the online layer is rarely a separate incident. It is the same conflict running continuously in a place the school cannot observe. (Final Report, October 2025, citing Bullying No Way 2024)
- ▸13% of all Australian children aged 10 to 17, not merely of those cyberbullied, said they had been told online to hurt or kill themselves, or that they should die. 7% within the past twelve months. (eSafety, 2024–25, n=3,454)
- ▸Among children who had been treated hurtfully online, disclosure fell with age: 64% of affected 14 to 17 year olds told a parent, against 75% of affected 8 to 10 year olds. (eSafety, Mind the Gap, 2021 fieldwork)
- ▸From Term 1 2027 Australian schools publish or link to anti-bullying policies aligned to the National Framework, or an equivalent jurisdiction framework. It expects schools to initiate a response within two school days of becoming aware of the behaviour. The clock starts at awareness. (Department of Education)
Where Guardii operates
Every figure above describes harm that occurs before any adult becomes aware of it. Guardii detects the behavioural pattern inside the private layer itself, reading sustained targeting, exclusion and coordinated pile-ons across the direct message and group chat layer where the survey data says most of it happens. It surfaces the pattern to a parent or a school while the child still has the option of saying nothing, and it reports patterns rather than transcripts so the child keeps their privacy. Where a national framework starts its clock at the moment a school becomes aware, detection is what moves that moment forward.
Overview
Our research combines data from multiple reputable sources — law enforcement agencies, child safety organisations, sports governing bodies, and academic studies — to present a clear picture of the online risks facing vulnerable populations. We examine online abuse targeting professional athletes and public figures alongside the rise in online grooming and sextortion targeting children. The findings inform the behavioural patterns the system detects at scale.
The AI inflection point
The single biggest shift since this page was last updated: generative AI has moved from an emerging risk to the primary accelerant across every category of online child harm. It has turned grooming into an automated, scalable activity, collapsed the time-to-harm in sextortion to a matter of hours, and put deepfake abuse within reach of anyone.
children had their images manipulated into sexually explicit deepfakes in a single year — in some countries 1 in 25 children.
UNICEF, ECPAT International & INTERPOL — Disrupting Harm Phase 2 (Feb 2026)increase in NCMEC CyberTipline reports involving generative AI in one year (4,700 → 67,000).
NCMEC / Thorn — 2024 CyberTipline Report analysisjump in generative-AI child-exploitation reports to NCMEC in six months (6,835 → 440,419).
NCMEC mid-year data (via HSToday)rise in AI-generated child sexual abuse videos found by the IWF (13 → 3,440) — 2025 was its worst year on record.
Internet Watch Foundation (Jan 2026)AI-generated images/videos of realistic child sexual abuse identified by the IWF in 2025 — 65% in Category A, the most severe.
Internet Watch Foundation — "Harm without limits" (Mar 2026)instances of grooming and sexual exploitation logged in just 50 hours of test chats with AI companion chatbots posing as minors.
Transparency Coalition (Oct 2025)AI is now the accelerant, not the exception
- ▸Scale without skill: AI tools have converged so abusive imagery can be generated with minimal effort and no technical expertise, removing the historic barrier to entry for offenders. (IWF, 2026)
- ▸Automated grooming at volume: offenders deploy AI chatbots that mimic a child’s speech, letting one perpetrator groom dozens of children simultaneously; NCMEC logged ~3,000 reports of people asking chatbots to assist with grooming or exploitation in 2025. (Bloomberg / NCMEC, 2026)
- ▸“Nudification” of real children: a significant share of new material is real children’s photos altered by AI, extending the harm to children who were never physically abused. (UNICEF, 2026)
- ▸On 19 January 2026 a coalition of UN bodies issued a Joint Statement on AI and the Rights of the Child, warning that society is collectively unable to cope with the scale of harmful AI-generated content. (UN News, 2026)
Where Guardii operates
Guardii is built for exactly this inflection point. Its real-time detection of grooming, sextortion, and AI-generated and deepfake child sexual abuse material runs inside children's direct messages across platforms — intercepting threats at the point of contact rather than removing material after it has already spread. As AI lowers the cost of abuse, pattern-based detection that scales with the threat is the proportionate response.
Online abuse against athletes & public figures
Professional athletes face unprecedented levels of online harassment, with abuse intensifying during major tournaments. Research from FIFA, FIFPRO, and World Athletics reveals systemic patterns of discriminatory behavior targeting players based on performance, race, and gender.
40+
languages covered by FIFA's Social Media Protection Service across 211 member associations
Source: Reuters90%+
reduction in athlete exposure to discriminatory content through automated moderation
Source: FIFAPlayer Abuse Rates at Major Tournaments
Percentage of players receiving discriminatory abuse on social media during tournament finals
Source: FIFA and FIFPRO studies on social media abuse during EURO 2020, AFCON 2021, and World Cup tournaments
Gender Disparity in Athlete Targeting
Women athletes face significantly higher rates of online abuse compared to their male counterparts
Source: FIFA World Cup data showing 29% higher targeting rate for women athletes
Key Findings
- •Discrimination is widespread: Over half of all players in major tournament finals experience discriminatory abuse, with racism being the most common form.
- •Women face disproportionate targeting: Female athletes are 29% more likely to receive abuse than male athletes, with 30% of abuse being sexualized in nature.
- •Automated moderation is effective: FIFA's Social Media Protection Service has hidden 2.6 million abusive comments and reduced athlete exposure to harmful content by over 90%.
- •Tournament intensity correlates with abuse: Abuse rates increase during high-stakes matches, with AFCON 2021 and EURO 2020 seeing over 50% of players targeted.
Protection at Scale
FIFA's Social Media Protection Service now covers all 211 member associations, monitoring content in 40+ languages and auto-hiding abuse before athletes see it. Independent studies confirm that automated moderation reduces athlete exposure to discriminatory content by over 90%, demonstrating that AI-powered protection systems can effectively shield public figures from the worst forms of online harassment.
Online abuse against women & influencers
Women with public online profiles—including influencers, content creators, journalists, and athletes—face dramatically higher rates of online abuse, harassment, and threats compared to the general population. Research from UNESCO, Pew Research Center, and eSafety Commissioner reveals systemic patterns of gender-based violence targeting women in public-facing roles.
73%
of women journalists have faced online violence, with 25% receiving physical threats and 18% sexual threats
Source: UNESCO1 in 3
women experience online abuse in a work context, with higher rates for those with public online/media profiles
Source: eSafety Commissioner3× Higher
Women are three times more likely than men to experience sexual harassment online (16% vs 5%)
Source: Pew Research Center36%
of women under 40 have received unsolicited sexual photos, with 67% arriving via social media
Source: YouGov7.1%
of tweets sent to women politicians and journalists are abusive or problematic
Source: Amnesty International84% More
Black women are 84% more likely to be targeted with abusive tweets than white women
Source: Amnesty InternationalSexual Harassment Rates by Gender
Women experience sexual harassment online at dramatically higher rates, with young women under 35 facing the highest risk
Source: Pew Research Center study on online harassment patterns
Key Findings
- •Public-facing women are disproportionately targeted: Women with public online or media profiles (creators, influencers, journalists) experience significantly higher rates of abuse compared to the general population.
- •Sexual harassment is pervasive: Women are three times more likely than men to experience sexual harassment online, with rates reaching 1 in 3 for women under 35.
- •Physical and sexual threats are common: Among women journalists who face online violence, 25% receive direct physical threats and 18% receive sexual threats—a pattern similar across public-facing roles.
- •Intersectionality matters: Black women are 84% more likely to be targeted with abusive content than white women, highlighting how racism and sexism compound online.
- •DMs are a primary attack vector: 67% of unsolicited sexual images arrive via social media, with Instagram reporting that high-profile women are "bombarded" with abusive content through direct messages.
Direct Message Risk
Instagram has implemented explicit-image blocking in DM requests after high-profile women reported being bombarded with abusive and sexually explicit content. The platform acknowledges that women with public profiles face persistent harassment through direct messaging, with cyberflashing (unsolicited sexual images) being a widespread problem. 67% of such images are sent via social media platforms, making DM protection essential for influencers and public figures.
Source: The Guardian
Key trends in online exploitation
Pandemic Catalyst
- COVID-19 lockdowns created a perfect storm for online exploitation
- Predators exploited increased online presence and decreased supervision
- Post-pandemic, exploitation rates continued to rise despite return to normal activities
Digital Platform Shifts
- Migration from public forums to private messaging
- Grooming cases now quickly move to private chats
- Increased use of ephemeral messaging apps that delete content automatically
Why These Trends Matter
Understanding these shifting trends is vital for creating effective protection systems. Guardii's technology specifically targets the private messaging environments where children are most vulnerable, using advanced AI to detect patterns that human monitoring might miss.
Global impact
Regional Analysis
North America
186K+ reports in 2023
Highest reporting capabilities with thousands of grooming and sextortion reports
FBI issued nationwide alerts due to "explosion" of sextortion cases
NCMEC reported a 132% increase in grooming cases from 2022 to 2023
Online exploitation is a global problem requiring coordinated international responses. While some regions have more robust reporting systems, all areas face significant challenges that require technological solutions like Guardii to detect and prevent harm.
Law enforcement responses
Law enforcement agencies worldwide are struggling to keep pace with the explosion in online exploitation cases:
Average Case Time
Time required for online grooming case investigation
Cross-Border Cases
Typical time to resolve international exploitation cases
Prosecution Rate
Of reported grooming cases result in successful prosecution
Sextortion Prosecutions
Even lower prosecution rate for sextortion cases
Resource Shortage
Of agencies report being understaffed for digital crimes
These statistics highlight the critical need for preventative measures like Guardii that can stop exploitation before it occurs rather than relying solely on after-the-fact law enforcement.
The global policy response
Within roughly a year, a wave of landmark legislation has landed — Australia's under-16 ban, the US TAKE IT DOWN Act, and the UK Online Safety Act — alongside a new UN normative baseline. These measures signal real political will, but most operate at the blunt end of the spectrum: account-level prohibition and mass age-verification.
Under-16 social media ban
World-first law in force 10 Dec 2025; by mid-January 2026 the government reported 4.7M+ under-16 accounts deactivated, removed or restricted, with platforms facing fines up to AUD $49.5M for systemic non-compliance.
DLA Piper / Wikipedia — Online Safety AmendmentTAKE IT DOWN Act
Signed 19 May 2025; criminalises non-consensual intimate images including deepfakes and requires covered platforms to remove reported content within 48 hours (full compliance by May 2026).
Congress.govOnline Safety Act — age checks
From 25 July 2025, sites must use “highly effective” age checks; Ofcom can fine non-compliant firms up to 10% of global revenue and pursue senior-manager liability.
OfcomJoint Statement on AI & the Rights of the Child
The UN bodies’ Joint Statement (19 Jan 2026) and UNICEF’s Guidance on AI and Children 3.0 (Dec 2025) set a global normative baseline for protecting children from AI-driven harm.
UN NewsPrevention is the gap these laws leave open
Every major authority — Europol, the WeProtect/Columbia University Global Threat Assessment, NCMEC, UN Women and UNICEF — now converges on the same conclusion: detection and removal after harm is failing to keep pace, and the decisive shift must be toward prevention at the point of contact, in the private messaging and gaming spaces where harm begins. Blanket bans restrict access for compliant families and push determined offenders to unmonitored channels; mass age-verification adds its own privacy burden. Guardii delivers the child-protection goal these laws pursue — without the collateral cost — by intercepting grooming, sextortion and abuse in real time, where it actually happens.
Grooming strategies & linguistic analysis
Research by Black et al. (2015) analyzing conversations between online predators and targets reveals several key grooming strategies:
Rapid Trust Development
Predators quickly establish trust through excessive compliments, showing interest in the child's problems, and providing emotional support.
Identity Deception
Appearing relatable by pretending to be younger, sharing similar interests, or creating false backgrounds to gain trust.
Sexual Desensitization
Gradually introducing sexual topics in conversation to normalize inappropriate discussions and content over time.
Isolation Tactics
Encouraging secrecy from parents and friends, positioning themselves as the only person who understands the child.
Boundary Testing
Consistently testing limits to gauge compliance levels and assess how far they can push the child.
The research highlights that online grooming is not strictly linear but often involves multiple tactics simultaneously. These insights inform Guardii's approach to detecting potentially harmful conversations.
Protection recommendations
Protection Recommendations
of exploitation occurs in private messaging
Focus on Direct Messaging
83% of exploitation occurs in private messaging environments, making these platforms critical for protection efforts.
Statistical evidence
Online Grooming Statistics
Grooming incidents have shown an alarming increase year over year, with experts concerned that reported cases represent only a fraction of actual occurrences.
*2024 and 2025 data projected based on current trends
Increase in reported cases since 2020
Reported grooming cases in 2023
Increase from 2022 to 2023
Projected cases for 2025
Estimated reporting rate (90% unreported)
Sextortion Statistics
Financial sextortion has emerged as a rapidly growing threat, particularly targeting teenage boys through social media and gaming platforms.
*2024 data projected based on current trends
Increase in financial sextortion (2022-2023)
Financial sextortion reports in 2023
Primary targets of financial sextortion
FBI sextortion complaints in 2020
Risk level projected for 2024
Why These Statistics Matter
The dramatic rise in both grooming and sextortion cases demonstrates the urgent need for better protection systems. Current safety measures are insufficient given the scale and growth of these threats.
Research sources
These sources have been carefully selected to provide a comprehensive and accurate overview of online child safety threats. The information and statistics used throughout this research page are derived from these reputable organizations and studies.