When the Association of Tennis Professionals (ATP) announced on 22 October 2026 that its AI-powered Safe Sport programme had prevented more than 234,000 abusive comments from reaching players and fans over the past year, the figure revealed more than the scale of online hostility in professional tennis.

It pointed to a potentially significant transformation in sport governance: artificial intelligence is moving beyond performance analysis and operational efficiency to become part of the infrastructure protecting athletes.

Similar developments are emerging across women’s tennis, international athletics and the Olympic movement. Together, they raise an important question: are sports organisations simply becoming better at filtering harmful content, or are they beginning to develop a new model of digital safeguarding?

ATP: From automated moderation to proactive protection

According to the ATP’s latest Safe Sport findings, the programme analysed more than 3.7 million comments across ATP and player social media channels during its second year.

More than 284,000 comments were identified as abusive, with over 234,000 prevented from reaching players and fans. More than 82% of the identified abusive comments were removed in real time.

The findings also indicated that more than 85% of players receiving social media engagement had been subjected to abusive comments originating from over 100,000 accounts.

Launched in 2024, ATP Safe Sport combines AI-enabled moderation with specialist threat assessment, investigations and cooperation with law enforcement authorities. Its technology partners include Tennis Data Innovations, Sportradar, Social Protect and Arwen AI.

As Reuters reported on 22 October, the initiative is part of a wider response to online harassment affecting professional tennis, including abuse linked to gambling and discriminatory attacks against players.

Importantly, ATP has expanded access to moderation services across men’s professional rankings, moving towards broader protection beyond the sport’s most prominent players.

The distinction between moderation and safeguarding is essential. Removing an offensive comment is a technological intervention. Assessing a credible threat, preserving evidence and coordinating an appropriate response are institutional responsibilities.

AI may accelerate the first process, but the second still requires human judgement, expertise and accountability.

Women’s tennis: What AI reveals about repeat offenders and gambling-related abuse

The ATP announcement becomes even more significant when considered alongside developments in women’s tennis.

On 16 July 2026, the Women’s Tennis Association (WTA) and World Tennis published their second annual report on online abuse directed at professional players and officials.

Their Threat Matrix initiative, delivered by Signify Group with specialist investigative partners, covers more than 10,000 players and officials. It combines AI-supported monitoring across 57 languages with human analysts, investigators and risk assessment specialists.

During 2025, the system analysed more than 1.2 million posts and comments, identifying over 12,000 abusive posts, including 3,726 categorised as serious abuse.

The International Tennis Federation’s reporting highlighted several important findings:

→ 66% of serious abusive posts or comments were removed.

→ 35 accounts associated with 12 individuals were escalated to law enforcement.

→ 89% of accounts responsible for serious abuse in 2024 did not reappear in 2025.

→ Just 9% of abusive accounts generated 87% of content classified as high concern.

These findings suggest that effective protection may depend less on detecting isolated offensive messages and more on recognising recurring patterns, concentrated harassment and potentially dangerous behaviour.

However, the absence of previously identified accounts in subsequent monitoring does not necessarily demonstrate that the individuals behind them stopped engaging in abuse. They may have changed accounts, platforms or methods.

This distinction illustrates why technological indicators require careful interpretation.

Perhaps the most striking finding concerns gambling.

According to the WTA and World Tennis report, angry gamblers were responsible for 42% of verified online abuse and 59% of serious abuse identified during the 2025 season.

This introduces an important dimension of sport integrity.

Online harassment is not necessarily an isolated social media problem. In some cases, it is connected to the wider commercial environment of sport, including betting markets and the behaviour of individuals reacting to sporting outcomes.

The involvement of Fanatics Sportsbook in investigative measures targeting abusive betting customers illustrates how safeguarding may increasingly require cooperation between sports organisations, technology providers, gambling operators and relevant authorities.

The future of athlete protection may therefore depend as much on institutional cooperation as on the sophistication of AI systems themselves.

World Athletics: From detecting abuse to understanding patterns of risk

The experience of World Athletics offers another perspective.

In December 2024, the organisation released findings from a four-year cyberbullying study covering the Tokyo Olympic Games, the World Athletics Championships in Oregon and Budapest, and the Paris Olympic Games.

As detailed in the report summarised by Sport Resolutions, more than 1.4 million posts and comments were analysed, with 2,438 athletes covered across the research programme.

The study documented persistent racism, sexism and sexualised harassment, alongside evolving methods of online abuse.

One particularly revealing finding concerned the increasing use of emojis to evade conventional moderation mechanisms.

This illustrates a fundamental challenge for AI-supported safeguarding: harmful communication is not static.

As detection systems improve, abusive behaviour can change in language, format and distribution. Algorithms therefore need continuous evaluation and adaptation, particularly when interpreting context, cultural differences and emerging forms of coded communication.

World Athletics also demonstrated the importance of moving from monitoring towards practical intervention.

Its programme included reporting abusive content to platforms, identifying threatening accounts, referring serious cases to authorities and providing year-round AI-supported protection to selected highly targeted athletes.

The broader lesson is that detecting harmful content has limited value unless organisations possess the procedures, expertise and resources to act on the information.

The Olympic movement: Digital safeguarding at a new scale

The International Olympic Committee has also developed AI-supported mechanisms for identifying online abuse directed at athletes.

During major Olympic events, these systems have combined automated monitoring with human assessment to identify potentially harmful content and support appropriate safeguarding responses.

Their relevance is particularly significant for international competitions, where athletes from different countries and linguistic backgrounds may experience online hostility across multiple platforms.

The Olympic context also raises questions about the protection of younger athletes, who may face additional vulnerabilities in digital environments.

For sports organisations, multilingual monitoring is becoming increasingly important. Yet language coverage alone does not guarantee accurate interpretation.

Context, cultural meaning, sarcasm, coded expressions and differences between criticism and harassment all complicate automated classification.

AI can help identify content requiring attention, but human expertise remains essential when determining whether a message represents a safeguarding concern or a credible threat.

This points towards a model in which technology supports professional decision-making rather than replacing it.

Beyond technology: Five questions for the future of AI governance

The growing use of artificial intelligence in athlete protection creates opportunities, but also introduces governance challenges that deserve greater attention.

1. Who defines harmful content?

Sports organisations need clear principles for distinguishing harassment and threats from legitimate criticism, public debate and expressions of disappointment.

Automated moderation should operate within transparent and proportionate frameworks that respect freedom of expression while protecting athletes from abuse.

2. How can human oversight remain meaningful?

AI systems can process enormous volumes of information, but decisions involving serious threats, investigations and referrals to authorities require accountable human judgement.

The objective should not be to eliminate human involvement, but to enable safeguarding professionals to concentrate their attention where it is most needed.

3. How should privacy and data protection be addressed?

Monitoring publicly available online content may still involve collecting, analysing and retaining information about identifiable individuals.

Responsible deployment requires attention to data minimisation, access controls, retention periods, lawful processing and appropriate oversight.

4. Should digital safeguarding become a standard component of athlete welfare?

The expansion of protection services across tennis rankings raises a broader question about equitable access.

Athletes outside elite competitions may also experience harassment, yet smaller federations and national organisations may lack the financial and technical capacity to operate sophisticated monitoring systems.

Shared services, partnerships and common safeguards could become increasingly relevant.

5. How should effectiveness be measured?

The number of comments detected or blocked is an important operational indicator, but it cannot independently establish whether athletes are safer.

More comprehensive evaluation could examine the accuracy of detection, response times, repeat-offender behaviour, athlete experiences, prevention of serious incidents and the effectiveness of referrals.

Without such measures, organisations risk confusing technological activity with demonstrated safeguarding outcomes.

The next frontier: From reactive moderation to preventive safeguarding

The emerging experience across sport suggests that AI-supported safeguarding may gradually move through three stages.

The first is detection: identifying abusive or threatening content at a scale that manual monitoring cannot reasonably achieve.

The second is response: connecting automated alerts with human assessment, investigations, platform interventions and support services.

The third, potentially more transformative stage, is prevention: identifying patterns of risk early enough to support timely interventions before online hostility escalates into more serious harm.

Future systems may become better at recognising coordinated harassment, identifying repeat offenders across digital environments, analysing emerging patterns of threats and helping safeguarding teams prioritise cases.

They may also support more consistent protection across languages, competitions and levels of sport.

These possibilities should not be confused with capabilities already proven to deliver measurable improvements in athlete safety. Predictive systems, in particular, require careful validation to avoid inaccurate classifications, disproportionate interventions or unjustified assumptions about individuals.

Nevertheless, the direction of development is increasingly clear.

AI is evolving from a tool that processes information into a technology that can support institutional risk management and athlete protection.

For sport governance, this creates an opportunity to establish common principles covering transparency, accountability, human oversight, privacy, proportionality and athlete participation.

It also suggests a growing role for cooperation between international federations, national sporting bodies, athlete representatives, researchers, technology providers and public authorities.

A new dimension of sport integrity

The significance of ATP’s announcement extends beyond the 234,000 abusive comments prevented from reaching players and fans.

It highlights a changing understanding of the environments in which sport integrity must be protected.

Athlete welfare is no longer confined to physical safety, competition conditions or traditional safeguarding procedures. Digital spaces have become part of athletes’ professional lives, public identities and exposure to risk.

Artificial intelligence can help sports organisations respond to this reality with greater speed and scale. But technology alone cannot establish trust, guarantee fairness or replace institutional responsibility.

The challenge for the coming years will be to transform technological capabilities into effective, accessible and accountable safeguarding systems.

The future of AI in sport integrity will not be defined simply by how much harmful content algorithms can detect, but by how effectively sport can turn that intelligence into meaningful protection while preserving human rights, fairness and institutional accountability.

That may be the most important lesson emerging from the ATP’s latest figures: the future of athlete safeguarding is increasingly digital, but its foundations must remain human.

 

IPSEI Perspective | This Featured Story presents an independent analysis of publicly reported developments in AI-supported athlete safeguarding. It draws on information published by sports organisations and other identified sources. The analysis and forward-looking observations are intended to contribute to constructive discussion on responsible innovation, sport integrity and institutional governance.