Emojis Targeting Black People Bundle

UPDATED APR 10, 2026
Today, October 16, 2025, the Oversight Board selected a case bundle appealed by Facebook and Instagram users regarding two pieces of content involving the use of monkey emojis to refer to Black people.
The first piece of content concerns a video posted on Facebook by a user in Brazil. The video features a scene from the movie in which two characters argue about ownership of a monkey. Text overlaying the video names the characters after European football clubs and refers to boys rising to prominence in Brazilian football, potentially implying that Brazilian athletes who are often Black are then recruited to European football clubs. The caption includes only a monkey emoji.
The second piece of content concerns a comment posted on Instagram in response to a video in which the posting user condemns a racist incident in Ireland they witnessed. The posting user described seeing a group of teenagers shouting a racist slur at a Black woman on the street and expressed heartbreak for the victim. The commenting user posted a statement expressing they do not support the posting user’s message and that they are eager for the situation to "blow up," and "to have some glorious fun with all the [monkey emojis] & out in the street."
Upon initial review, Meta left this content up. However, upon further review, we determined the content did in fact violate our policy on Hateful Conduct, as laid out in the Community Standards. We therefore removed the content.
Under our Hateful Conduct policy, Meta prohibits content “targeting a person or group of people …on the basis of their … protected characteristic(s),” including race, ethnicity and national origin, “in written or visual form.” This encompasses “dehumanizing speech in the form of comparisons to or generalizations about … animals in general or specific types of animals that are culturally perceived as inferior.” These comparisons can be shown visually through the use of emojis and discerned by the content’s context.
We will implement the Board's decision once it has finished deliberating, and will update this post accordingly. Please see the Board's website for the decision when they issue it.
Read the board's case selection summary
Case decision
We welcome the Oversight Board’s decision today, February 10, 2026, on this case. The Board overturned Meta’s original decision to leave up both pieces of content. Since Meta previously removed the posts, there will be no further action on the case content.
When it is technically and operationally possible to do so, we will also take action on content that is identical and in the same context as the first case. For more information, please see our Newsroom post about how we implement the Board’s decisions.
After conducting a review of the recommendations provided by the Board, we will update this post with initial responses to those recommendations.
Recommendations
Recommendation 1 (Implementing in Full)
To improve the ability of its automated systems to more accurately detect the use of emojis in violative contexts, Meta should audit its training data used for Hateful Conduct policy enforcement and ensure the data is updated periodically to include examples of content with emojis in all languages, violating use of emojis and new instances of the hateful use of emojis.
The Board will consider this recommendation implemented when Meta provides the Board with detailed results of its first audit and the necessary improvements that the company will implement as a result.
Our commitment: Our Hateful Conduct classifier training data includes emojis across languages, including examples of the hateful use of emojis. Our training data is generally updated on a daily basis and incorporates content from user reports to ensure that it remains current and comprehensive. We also continuously monitor the precision of our Hateful Conduct classifiers and investigate trends as they arise to maintain accuracy. We will consider how to share relevant insights on how we’ve assessed our training data in a future confidential update to the Board.
Considerations: We maintain transparency about how our enforcement technology operates from the development of new content enforcement tools to the training of classifiers that detect specific signals and enforce across all our policies. Training classifiers is an iterative process, as this technology learns from thousands of human decisions to increase accuracy over time. These classifiers are principally used to assess content reported by users and are continuously improving in accuracy and scope. As our policies evolve, we review and update our training data to ensure alignment with policy scope, including the Hateful Conduct policy. During events such as the World Cup, where we experience temporary spikes in violating content or user reports, we may utilize targeted proactive measures where warranted, to reduce exposure to high volumes of violating content. The classifiers are trained on diverse datasets of sampled content that include text and emoji-based examples of Hateful Conduct violations.
Classifiers are one part of how we approach automated enforcement. Recently, we announced that more advanced AI will also be leveraged to improve our content enforcement systems, particularly because they can better adapt to understand cultural nuance, niche subcultures, and rapidly changing or regionally specific code words, emoji meanings, and slang. For the time being, our classifiers will be used in tandem with the advanced AI and human review. Furthermore, classifier training data will include samples of content that has been enforced on by our advanced AI. Metrics in future iterations of the Community Standards Enforcement Report and other regulatory reports will be inclusive of advanced AI enforcement. We will provide updates in future reports to the Board as we consider ways to share further insights on our classifier training data confidentially with the Board.
We currently have a number of ways of addressing content that targets both private and public figures through either our Bullying and Harassment policy or our Hateful Conduct policy, which we detail in the considerations below. As part of these efforts and as detailed further in recommendation 3, we also have a number of tools and initiatives to address “campaigns” through both direct and indirect means, which we believe complete this recommendation as Work Meta Already Does.
Our commitment: We currently have a number of ways of addressing content that targets both private and public figures through either our Bullying and Harassment policy or our Hateful Conduct policy, which we detail in the considerations below. As part of these efforts and as detailed further in recommendation 3, we also have a number of tools and initiatives to address “campaigns” through both direct and indirect means, which we believe complete this recommendation as Work Meta Already Does.
Considerations: While our policies do not explicitly call out “campaigns”, we do have a number of policies and enforcement mechanisms as part of our Bullying and Harassment Community Standard and our Hateful Conduct Community Standard that allow us to address content at scale and on-escalation that may target specific individuals and groups of people based on their protected characteristics.
As the Board notes in its decision, under our Hateful Conduct Community Standards we remove direct attacks that contain dehumanizing speech in the form of comparisons to or generalizations about animals or specific types of animals that may be culturally perceived as inferior when targeting individuals or groups of people on the basis of their protected characteristics. This includes comparisons between Black people and apes or ape-like creatures. These targets can be identified by reviewers through textual as well as visual signals, and these comparisons can include visual comparisons through the use of emojis.
Under our Bullying and Harassment Community Standard, we also outline how we approach various potential targets, with specific protections for everyone and additional protections for private individuals, limited scope public figures, and minors. This includes protections against dehumanizing comparisons, where we distinguish between private individuals and public figures to allow additional commentary on people with a large public audience. The full details on the specific protections for each group are outlined in our Transparency Center.
Finally, on escalation, we remove targeted mass harassment, including when any individual is targeted through personal surfaces, such as through their inbox or profiles, with either content that violates our Bullying and Harassment policies for private individuals, or objectionable content that is based on a protected characteristic under our Hateful Conduct policy. We may also remove directed mass harassment, when the content targets "individuals at heightened risk of offline harm” on any surface (e.g., human rights defenders or opposition figures in at-risk countries during election periods). Finally, we may remove accounts that engage in mass harassment if they are a part of (a) state or state-affiliated networks targeting any individual via any surface or (b) adversarial networks targeting any individual via any surface with content that violates the Bullying and Harassment policies for private individuals, content that targets them based on a protected characteristic under our Hateful Conduct policy, or content or behavior that may considered objectionable with local context from internal experts.
We provide more details on our approach to this type of content in the context of global events below. To ensure these policies are applied appropriately including in cases where there may be contextual ambiguity, we provide feedback to reviewers and continuously work to improve our systems. As a result, we consider this recommendation fulfilled as work that Meta already does and will have no further updates.
To ensure its systems are well-equipped to address hateful campaigns during major sporting events, such as the FIFA World Cup, Meta should ensure that its time-sensitive mitigation efforts, be that through its Integrity Product Operations Center or another risk mitigation system, include active monitoring of content with emojis that incite targeted discrimination or hostility in the lead up, during and in the immediate aftermath of these events.
The Board will consider this recommendation implemented when Meta shares with the Board evidence confirming the deployment of its risk evaluations and mitigation efforts used during major sporting events.
Our commitment: We regularly support time-sensitive efforts to reduce and mitigate harm during major sporting events like the FIFA World Cup, including measures to address content that may incite discrimination or hostility. These efforts involve a highly coordinated effort across the company to actively monitor for things like the use of emojis to target individuals or groups in ways that violate our policies. We are committed to continuing these efforts for major sporting events, and will share more about our efforts related to the 2026 FIFA World Cup as we work to ensure that risk evaluations and mitigation efforts are deployed effectively during the event.
Considerations: Ahead of major sporting events like the FIFA World Cup, we establish dedicated working groups to identify and safeguard the authentic accounts of athletes, federations, and key participants. Our operations teams leverage historical data and insights to identify the right approach to monitoring and enforcement for each event. Where needed, they may determine that proactive measures should be used to identify content or comments that violate our Hateful Conduct policy before users report them. These measures may be launched before, during, and after the event, and may also be tailored to multiple languages and event-specific risks.
As indicated by the Board in their decision and in our Bullying and Harassment Community Standard, we do not allow content calling for, or stating an intent to engage in behavior that would qualify as, bullying and harassment under our policies. When we have confirmation from the target or an authorized representative of the target that the content is unwanted, we will remove this content. In the case of an event like the World Cup, athletes that are enrolled as managed partners can identify a range of people as “authorized representatives”—such as federations, managers, and clubs—to submit formal reports or escalations through our support channel for managed partners, which is known as the Media Support Portal.
As part of our standard approach to major events, the working group conducts post-event reviews to track results and impact and then communicates their findings to internal stakeholders to inform future improvements to our mitigation efforts. Our playbooks and processes are continuously updated based on event learnings, ensuring our approach evolves with emerging risks. We will provide an update to the Board in a future report regarding our efforts around the upcoming FIFA World Cup, where we intend to deploy these measures.