As the annual 16 days of activism against gender-based violence draw to a close today December 10, which is also International Human Rights Day, one thing remains painfully clear. Telling the truth is still risky for many people, even online. Survivors try to describe what happened to them, yet the platforms they depend on often punish them for using the right words.
This tension frames the digital age we live in. Social networks that began as casual spaces for entertainment and connection have become central to public life. They now carry the weight of testimony, activism, and community building. People use them to assert authorship over their own stories, especially those long denied the right to speak openly. Yet as these digital commons expand, their freedom contracts. The systems once designed to widen expression are increasingly mirroring offline hierarchies and policing language in ways that silence the very people who most need to be heard.
It is this contradiction that has pushed many users to ask why the word rape is now being replaced with grape on social media platforms. The trend is everywhere. The word itself is ugly for an even uglier act, and that is precisely why many insist it must be named clearly and shamed as it is, without dilution. Yet naming it now attracts backlash from automated systems that cannot interpret context. Posts are flagged, shadow banned or deleted not for promoting violence but for identifying it. Many users across the world have recounted how they received warnings simply for sharing testimonies of sexual or gender-based violence. This global digital experience is now forcing individuals to twist language in order to escape algorithms and remain visible online.
Rather than risk blowback, users, especially women already navigating precarity now misspell rape, break it with asterisks, substitute the description with clusters of grapes or purple emojis, or erase their stories completely for fear of losing cherished pages. This evasive vocabulary belongs to a growing lexicon known as algospeak, a coded language used to avoid sanctions on social media platforms.
But how did we get here? Content policing grew out of the early chaos of social media, when platforms became fertile ground for graphic violence, hate campaigns, extremist recruitment, and coordinated harassment. Under pressure to mount stronger oversight, companies like Meta started building systems to detect and remove harmful content. But the scale was impossible. Millions of posts appear every minute, far beyond what the existing human workforce could reasonably review. A report by The Guardian once exposed that a 4,500-member review team at Facebook was tasked with handling “more than 100 million pieces of content every month,” leaving moderators roughly ten seconds to make a judgement call on each piece.
Faced with this impossible pace, companies embraced automated systems that promised the speed and scalability needed to keep up with enforcement.
Machine Eyes, Human Costs
What is now known as social media algorithms when it comes to content moderation are essentially sets of instructions that allow computers to identify patterns and make predictions. They are trained on large datasets to detect indicators of harm and operate through classification, a process that sorts content into categories such as safe or unsafe. They are powerful and quick, yet unable to grasp nuance or intention, and this is where the problem begins. The result is a machine logic that collapses radically different types of speech into a flattened zone of suspicion.
Even Meta, in a statement that promised “more speech and fewer mistakes,” conceded that its automated systems often “get things wrong” because they cannot interpret context in the way human beings do, and that its appeal processes are “frustratingly slow” and “do not always get to the right outcome.” In response to this problem, the tech giant is now shifting to a “Community Notes” model — an approach also used by X (formerly Twitter) that transfers some responsibility to users by allowing them to annotate content.
The failures of algorithmic judgement, however, cannot be separated from the economics that drive Big Tech. Instead of strengthening human moderation, companies have thinned out the workforce, outsourcing and then downsizing the very people who can read and interpret human testimonies. Earlier this year, TikTok laid off hundreds of moderators in favour of AI systems that regularly misclassify trauma testimonies, satire, minority languages, and basic news. The erosion of human oversight consolidates the authority of systems whose interpretive imagination is painfully limited.
At the end of the day, these limitations fall hardest on communities whose survival depends on context being understood. Nowhere is this clearer than in Africa and other regions marked by deep gender inequality and social volatility. The trend I now refer to as the fruitilisation of rape, which manifests online as the translation of unspeakable crime with sweetened symbols is one of the clearest illustrations of how restrictive automated actions can become instruments of harm. This linguistic distortion trivialises violence and compounds the emotional burden of survivors in societies where rape is already chronically underreported due to stigma, threats, and slow judicial processes.
Worse still, misogynists and digital misfits who understand that euphemised vocabulary provides cover for harmful rhetoric now exploit this vacuum, bandying the grape metaphor as jokes or threats and reproducing violence in spaces where survivors are penalised for naming what happened. If this is the reality, then the conclusion must be honest. There can be no sweet rape, and the task is not to invent gentler metaphors but to dismantle the architecture that demands them.
What Must Change
Meaningful reform requires a major shift in what platforms understand as their core obligation. Without a reset that centres user safety, narrative dignity, and democratic accountability, all improvements will be cosmetic. Repairing the harm demands restoring human judgement as the backbone of moderation. This means retaining a global workforce of trained, adequately supported moderators who can exercise careful judgement without impossible quotas. Machines may assist, but they cannot be entrusted with moral authority. Strengthening labour protections, especially for outsourced moderators in the Global South whose rights are steadily eroded by layers of exploitative subcontracting practices, is also central to this task.
Redistributing editorial power is equally urgent. AI moderation cannot be built on datasets drawn solely from the Global North. Platforms must integrate the perspectives of user collectives, survivors, rights advocates, and affected communities into the design of datasets, context guidelines, and algorithmic thresholds. A humane digital environment must also protect testimony as a right. Users should be able to tag content to indicate that they are naming violence, not promoting it, preserving clarity of language against literal minded deletions. As platforms turn to community driven moderation systems, the shift holds promise because it restores interpretive authority to users. But without safeguards, it risks replicating the same patriarchies, prejudices, and moral policing that already silence survivors. Any move toward community led oversight must therefore be transparent, globally representative, and reinforced by strong safeguards that prevent hostile majorities from deciding whose testimonies are recognised as legitimate.
