Watching the Feeds: AI Monitoring of Online Abuse Against Athletes

Timeline

1. Jul 2021World Athletics monitors abuse at the Tokyo Olympics; 63% of identified abuse targets two female athletes.
2. Jul–Aug 2023FIFA and FIFPRO's Social Media Protection Service covers the Women's World Cup.
3. Jan 2024The WTA and ITF launch AI monitoring covering about 8,300 players.
4. Jul–Aug 2024World Athletics and the IOC run monitoring at the Paris Olympics.
5. Oct 31, 2024World Athletics publishes its Paris study, with 809 verified abusive posts.
6. Dec 18, 2024World Athletics publishes a four-year analysis of abuse in athletics.
7. Jun 2025The WTA and ITF report on abuse in 2024, finding 40% came from angry gamblers.

The ethical issues

• Duty of care owed to athletes facing online abuse.
• Gendered and racist abuse and how it is measured.
• Privacy and consent in monitoring athletes' accounts.
• Limits of AI and the need for human review.
• The responsibility of platforms and of betting markets.

Stakeholders

Stakeholder	Interests & obligations
Athletes, especially women and minority athletes	Safety, mental health and freedom to use social media.
Sports governing bodies	Athlete welfare and the reputation of their events.
Monitoring firms	Accurate detection and business contracts.
Social media platforms	Content moderation and their own policies.
Law enforcement	Pursuing criminal threats.
Fans and bettors	Free expression within limits.

The case

For many athletes, the worst part of a loss now arrives on their phones. Strangers send insults, slurs, sexual comments and threats, often within minutes of a result. Women athletes face a particular mix of sexualized and sexist abuse. Since 2021, sports bodies have turned to artificial intelligence to fight back, hiring firms to scan millions of posts, flag abuse for human review and report the worst to platforms and police. The programs have produced the first large-scale evidence of who is targeted and by whom. They also raise questions about what sport owes its athletes online, and what monitoring can and cannot do.

Background

World Athletics was an early mover. Working with Signify Group’s Threat Matrix service, it monitored athletes’ accounts at the Tokyo Olympics in 2021. Of the 161 athletes covered on Twitter, two, both women, received 63% of the abuse identified; 29% of the abuse was sexist and 26% racist. At the 2022 World Championships, 59% of abuse targeted female athletes. At the 2023 Championships, 35% was racist.^[2]

Abuse is not limited to women, and patterns vary by event. At Paris 2024, the IOC said gender-based abuse was the dominant type, with sexual and sexist abuse aimed at female athletes, while about 70% of all targeted athletes were male.^[5] Football has seen high-profile racist abuse of men, as in the case of Euro 2020: Online Racist Abuse of England’s Penalty Takers.

The rules and governance

No single rule requires sports bodies to protect athletes online. Their duty comes from safeguarding policies and a broader duty of care. World Athletics, for example, has required its member federations to adopt safeguarding policies.^[1] Platforms enforce their own content rules, and threats can be crimes under national law. Monitoring services work between these systems: AI flags posts, human analysts verify them, and the sports body decides whether to report to platforms, share with security teams or refer to police.^[4]

Monitoring at the Women’s World Cup

FIFA and the players’ union FIFPRO ran a Social Media Protection Service at the 2023 Women’s World Cup, from July 19 to August 21. It analyzed 5.1 million posts and flagged 102,511 for review; analysts verified 7,085 as abusive, and the service hid 116,820 comments from players’ feeds. It covered 697 players and coaches; more than 150 of them, about one in five, were targeted. The share of homophobic abuse was almost twice that recorded at the men’s World Cup in Qatar in 2022.^[3]

The report described limits. Monitoring could run without any action by players, but moderation, which hides abusive comments, required each player to opt in. On X the service could not fully hide replies, and TikTok offered no way to moderate through its interface.^[3]

Tennis and the gamblers

In January 2024 the WTA and ITF launched Threat Matrix for about 8,300 players, who were covered automatically. In 2024 it analyzed 1.6 million posts and verified about 8,000 as abusive, violent or threatening, from 4,200 accounts. Of the 458 players targeted, five received 26% of the abuse. Angry gamblers sent 40% of all detected abuse, and 77% of the abusive direct messages that players reported. Fifteen cases were referred to law enforcement, three of them to the FBI.^[4] An interim count for January to October, a period that included two Grand Slams, had put the gamblers’ share at 48%.^[4]

The WTA and ITF called on the gambling industry to act. Player council member Jessica Pegula said: “Online abuse is unacceptable, and something that no player should have to endure. … But it’s not enough on its own. It’s time for the gambling industry and social media companies to tackle the problem at its source.”^[4] Pegula had earlier described “constant death threats” and “family threats” as being “normal now.”^[6] Signify’s chief executive, Jonathan Hirshler, said “a relatively small number of accounts are responsible for a significant proportion of prolific abuse.”^[4] Betting-related harassment of college athletes is covered in College Athletes, Player Prop Bets, and Harassment by Bettors.

Paris 2024

At the Paris Olympics, World Athletics monitored 1,917 athletes and officials across X, Instagram, Facebook and TikTok. Of 355,873 posts in 36 languages, the AI flagged 34,040 for human review, and 809 were verified as abusive. Sexualized abuse made up 30% and racism 18%; two athletes received 82% of the abuse. Female athletes faced a mix of sexual and sexist abuse.^[1] The IOC ran its own service for about 10,400 athletes and officials, scanning some 2.4 million posts in more than 35 languages. It flagged more than 152,000 as potentially abusive and verified more than 10,200, from 8,900 accounts, aimed at 353 athletes and officials. “These figures demonstrate the scale of protection that was provided … and the scale of online abuse that athletes and officials face,” said Kristy Burrows, the IOC’s head of safe sport.^[5]

Aftermath

World Athletics’ four-year analysis, published in December 2024, covered more than 1.4 million posts and 2,438 athletes. It noted that the study excluded direct messages, “suggesting the issue is even greater than the study highlights,” and that abusers increasingly used emojis to evade platform rules. Over the period it reported 1,258 posts to platforms, escalated 254 accounts, offered year-round AI protection to 25 highly targeted athletes and referred two serious cases to authorities. It said it was consulting athletes on future protection.^[2] Valerie Adams, chair of its Athletes’ Commission, said many athletes “choose to switch off their accounts during those major milestone moments.”^[2] The WTA and ITF said they would add moderation to hide abuse in real time.^[4]

The core dilemma

Athletes need social media for their careers, sponsors and fans, yet it exposes them to abuse that harms their mental health. Sports bodies cannot control platforms or the people who send abuse, but they can monitor, report and support. AI makes that possible at scale, though it misses private messages and cannot stop abuse at its source. How far should sport go in watching over athletes online, and who else must share the responsibility?

Competing values

• Protection vs. privacy and autonomy: shielding athletes from harm against their control over their own accounts.
• Scale vs. accuracy: AI’s reach against the need for human judgment.
• Free expression vs. safety: criticism of athletes against abuse and threats.
• Shared responsibility vs. passing the buck: sports bodies, platforms, betting firms and police.

Strongest arguments

For AI monitoring. The volume of posts is too large for humans alone. Monitoring documents the problem, identifies prolific abusers, hides harmful content and supports referrals to police and venue bans. The data show that a small number of accounts and a few targeted athletes account for much of the abuse, allowing focused protection. It fulfills part of a sports body’s duty of care.

For caution. Monitoring can be intrusive, and athletes may not know or agree to how their accounts are scanned. AI can misread context, which is why many flagged posts are not verified as abuse. Programs that rely on platforms’ cooperation have limited reach, and most private messages are excluded. Monitoring may also let platforms and betting companies avoid their own responsibility.

For a broader approach. Monitoring is one tool among several: platform design, betting industry controls, education, mental health support and law enforcement all play a part. Pegula’s call to tackle “the problem at its source” reflects this view.

Applying the frameworks

Care ethics

Care ethics would ask whether athletes, especially young women and those targeted most, feel supported. It would value giving athletes control, such as opting in to moderation, and pairing monitoring with mental health support, as the IOC did. It would ask how organizations respond to individual athletes, not just to aggregate numbers.

Utilitarian

A utilitarian would weigh the harm prevented, such as hidden posts, removed accounts and threats referred to police, against the costs of monitoring, including errors and privacy intrusion. With harms concentrated on a few athletes and abusers, targeted protection may produce large benefits for modest cost. The utilitarian would also count the benefit of pressure on platforms and betting firms.

Rights

A rights lens would protect athletes’ right to safety and to participate in public life without harassment, while respecting their privacy and the free expression of others. It would favor athlete consent and transparency about how data are used, and clear lines between criticism and abuse.

Walk it through: the Simple Decision Model

1. Spot the issue. You lead athlete welfare at a governing body planning protection for a major event. Stakeholders include athletes (especially those most often targeted), coaches, the monitoring firm, platforms, betting operators, police and fans. The codes in play are safeguarding policies, platform rules, data protection law and criminal law.
2. Gather facts. Who has been targeted at past events, on which platforms and by whom? What can the service see and not see, such as direct messages? Watch for metric bias: counting flagged posts can look like success without showing whether athletes feel safer.
3. Test options. Monitor all athletes automatically; offer opt-in moderation; focus on highly targeted athletes; or combine these with education and mental health support. Run each through care, consequences and rights.
4. Check codes and law. Confirm data protection compliance, platform terms and how threats are referred to police.
5. Decide and document. Explain what is monitored, how, by whom and with what safeguards. Consult athletes, as World Athletics said it was doing.
6. Communicate and follow up. Tell athletes how to report abuse and get support, publish results after events, and press platforms and betting firms on their roles.

What’s settled and what’s contested

Settled

• World Athletics, FIFA and FIFPRO, the WTA and ITF, and the IOC have used AI with human review to monitor abuse of athletes.
• These programs found that a few athletes and accounts receive or send much of the abuse, and that female athletes face sexualized and sexist abuse.
• The WTA and ITF found that angry gamblers sent 40% of detected abuse in 2024.

Contested

• Whether monitoring meaningfully reduces harm or mainly documents it.
• How much responsibility rests with platforms and betting companies.
• How to balance protection with athletes’ privacy and control.

Why this case matters

Online abuse is part of the modern athlete’s working environment. These programs show that sports bodies can do more than condemn abuse, but also that technology alone will not solve it. The case is a test of what duty of care means in a digital world: protecting people where they work, even when the workplace is someone else’s platform.

Key terms

Duty of care

The responsibility of coaches, organizations and institutions to protect the health, safety and well-being of those in their charge. It is not a license for total control.

Non-maleficence

The duty to avoid causing harm, often paired with care ethics when setting safety rules.

Stakeholders

Everyone with something at stake in a decision: athletes, coaches, parents, officials, fans, employees, sponsors, media, communities and regulators, including the less visible ones.

Controls

Concrete safeguards that make ethics operational: training, disclosure forms, testing, monitoring, hotlines. Ethics needs principles, process and controls.

Consent and autonomy

Whether people have agreed freely and with full information, and whether their right to make their own choices is respected.

Institutional accountability

Structures that make a sport organization explain its decisions and face consequences: clear authority, transparency, appeals, independent oversight and enforcement.

Definitions from the Ethics Cases glossary.

Discussion questions

1. What duty of care does a governing body owe athletes on social media?
2. Should monitoring of athletes' accounts require their consent? Why or why not?
3. The WTA said 40% of abuse came from angry gamblers. What responsibilities do betting companies and sports bodies have?
4. AI flagged tens of thousands of posts, but humans verified far fewer. Why does human review matter?
5. What can platforms do that sports bodies cannot?

Sources

1. Study on online abuse of athletes at the Paris 2024 Olympic Games — World Athletics
2. Four-year analysis of online abuse in athletics — World Athletics
3. FIFA/FIFPRO Social Media Protection Service report: FIFA Women's World Cup 2023 — FIFPRO
4. WTA and ITF publish season-wide online abuse and threat report — WTA
5. How the IOC protected Paris athletes from cyberbullies — Deseret News
6. Social media abuse in tennis: Gamblers send 48% of offensive messages — BBC Sport