UpScrolled’s breakneck rise is colliding with the hard reality of content safety. After a surge in sign-ups, the young social network is struggling to remove usernames and hashtags that contain slurs, and users say harmful posts are staying up for days. The result is a credibility test for a platform that promises “equal power” for every voice but is now under pressure to prove it can police hate speech at scale.
The app’s growth accelerated amid upheaval at a larger rival, pushing UpScrolled past 2.5 million users in January, with more than 4 million cumulative downloads since mid-2025, according to Appfigures. That momentum appears to have outpaced the company’s enforcement systems: reports describe slur-laden handles and trending tags that weren’t actioned even after being flagged, a visible sign that moderation queues and automated filters are falling behind.
Rapid Growth Outpaces Safety and Moderation Systems
Early-stage social networks often underestimate how quickly abuse scales. The Integrity Institute has noted that fast-growing communities commonly see report volumes multiply overnight during migration waves, stressing both machine classifiers and human review. Large platforms now aim to remove the majority of violating content proactively; Meta has reported proactive detection rates above 90% for hate speech in recent transparency reports. New entrants typically start far below that threshold, relying heavily on user flags and creating visible lag.
UpScrolled’s own policies mirror industry norms—prohibiting hate speech, harassment, and content intended to cause harm—but policy text is only as strong as the systems enforcing it. The most glaring misses so far are in “creation-time” safety: blocking slurs in usernames, profile fields, and hashtags before they go live. Those are low-friction vectors for abuse and should be guarded by strict dictionaries, obfuscation detection (e.g., leetspeak), and language coverage across dialects and slur variants.
Visible Gaps in Enforcement Undermine Trust
Users have documented accounts whose handles include racial and other hate slurs as well as tags organized to brigade targets—classic signals that can be caught without deep linguistic context. Persistent visibility of those assets suggests missing guardrails: insufficient blocked-term lists, weak fuzzy matching, or inadequate human triage to sweep through high-severity reports. When violations are obvious and remain live, trust erodes quickly, and bad actors learn the system can be gamed.
Compounding the issue, hate speech evaders mutate spellings and embed coded language to dodge simple keyword screens. Effective systems loop user reports into model training pipelines, use active learning to capture new variants, and routinely re-scan old content as classifiers improve. That continuous-improvement cycle appears to be in its infancy at UpScrolled, leaving users to shoulder the burden of flagging—and waiting.
Company Response and Plan to Improve Moderation
In a public video statement, CEO Ibrahim Hijazi acknowledged that “harmful content” has been uploaded in violation of UpScrolled’s terms and said the company is rapidly expanding its content moderation team while upgrading its infrastructure to catch and remove violations more effectively. In messages to users and reporters, the company has also advised people not to engage with bad-faith actors while it scales enforcement.
That playbook—hire, harden, and automate—tracks with how other platforms have rebounded from safety lapses. But execution speed matters. Effective triage typically prioritizes creation-time blocks (usernames, bios, hashtags), automated takedowns for high-confidence hate speech, and 24/7 coverage for escalations. Publishing a near-term service-level target (for example, removing high-severity hate content within hours) can also give users and advertisers a concrete yardstick.
Lessons From Other Platforms on Handling Hate Speech and Safety
Bluesky faced a similar wave in 2023 when slur-based usernames slipped through, prompting backlash and a swift tightening of handle policies. Larger platforms mix multiple defenses: pre-publication screening; contextual models to reduce false positives against reclaimed or colloquial terms; rate limits and friction for newly created accounts; and community tools such as crowdsourced notes to add context. None of these alone solves hate speech, but together they raise the cost of abuse and compress the window of harm.
Wider context underscores the stakes. The Anti-Defamation League’s most recent Online Hate and Harassment report found that more than 50% of U.S. adults experienced harassment online, and roughly a third described incidents tied to hate. That reality means migration waves don’t just bring fresh users—they bring the full spectrum of online behavior, from healthy debate to organized abuse.
What It Will Take to Rebuild Trust on UpScrolled
UpScrolled can demonstrate traction quickly with a few high-impact shifts: enforce a zero-tolerance blocklist for handles and tags at creation; add fuzzy and multilingual matching; deploy real-time model checks on posts; and staff a round-the-clock escalation lane for hate content. Publishing a short, frequent transparency snapshot—proactive vs. reactive removal rates, median takedown times, and appeal outcomes—would show whether detection is catching up.
Advertisers are watching, too. Many evaluate platforms against brand safety frameworks such as the Global Alliance for Responsible Media and expect clear metrics. Hitting aggressive but public goals—say, >80% proactive detection in the near term, trending higher—would signal that UpScrolled’s safety systems are maturing alongside its user growth.
The company’s promise of open expression resonates, but a functioning baseline against hate speech is non-negotiable. Right now, the gap between policy and practice is the story. How quickly UpScrolled closes it will determine whether the platform’s viral moment becomes a durable community—or a cautionary tale.
