Apple Peiqing Ni was tagged in twelve deepfake posts portraying her as a drug addicted sex worker after a Tiananmen memorial post. X's automated review and human support both said the content was fine. Only a Guardian inquiry moved the platform to suspend the account.
When Apple Peiqing Ni told X's automated moderation system that someone had tagged her in twelve deepfake posts calling her a drug-addicted sex worker, the platform said the content did not breach its rules. When she appealed to a human support agent, she got the same answer. The account that ran the campaign stayed live until the Guardian raised the case with X's press office.
Ni, 27, runs the China Dissent Network, a UK-based group that supports Chinese dissidents abroad. She moved to London in 2019 to study and has spent years posting about human rights in China, including the 1989 Tiananmen Square crackdown. In early June, she shared a poster saying she would join a Tiananmen commemoration in Sutton, south London, on 4 June. Within hours, the campaign began.
The twelve posts tagged Ni and used synthetic images and videos. Captions described her as having "chronically chaotic sexual relationships" and being a heavy drug user. One post falsely claimed, and celebrated, that Ni had been "beaten badly on the red streets of London while protesting with other anti-China groups," calling it "the perfect retribution for her extreme behaviour." The reference appeared to be a March attack by masked men on a male activist at a Trafalgar Square sit-in that Ni had organised. Ni had not been assaulted.
The contrast with X's own written rules is sharp. The platform prohibits, in its published harassment policy, "the malicious, unreciprocated targeting (such as mentioning or tagging) of individual(s), particularly when shared to humiliate or degrade someone." Ni's complaint quoted the rule back. The automated system still concluded the posts fell outside it. A follow-up complaint through X's support service was also rejected.
X declined to explain to the Guardian why the automated system and the human reviewer both cleared the posts. After the inquiry, the offending account was suspended. Ni was later told the action had been taken in response to "different reports" about the content, not the ones she had filed.
UK police had earlier told Ni to take the matter up with the US-headquartered platform rather than pursue a UK investigation. That left her with X's internal systems as her only practical remedy. The cascade that followed, automated rejection, human rejection, and action only after a journalist's call, is the system a high-risk user actually faces.
Ni believes the campaign was run by a pro-regime bot. The Guardian's reporting treats that as her attribution, not a confirmed link to any operator. Whoever ran the account exploited a familiar gap: rules that read as protective on a help page, paired with an enforcement pipeline that fails to apply them. The case sits alongside a long pattern of complaints from dissidents, journalists, and women targeted by synthetic sexual imagery, in which a platform's stated policy and its actual response diverge.
The constructive question is what an escalation path for users like Ni would look like in practice. Her case was resolved only because a major outlet made a phone call. Most users in her position do not have that option. A named human-review lane for accounts reporting targeted synthetic-media abuse, a published turnaround time, and a public log of policy citations used to clear or uphold complaints would each give the rule a chance of being the rule.
Watch whether X publishes any explanation of why the posts cleared its systems, whether other dissidents in similar positions have had the same experience, and whether the platform updates its stated turnaround for what it calls targeted harassment complaints.