OpenAI bans Russian op “Stop News” that used AI to pose as UK and African media

OpenAI has pulled the plug on a Russian influence operation that used its AI models to churn out content masquerading as legitimate news outlets in the UK and Africa. Dubbed “Stop News,” the cluster of accounts was banned after the company cross-referenced off-platform indicators from Meta’s Q2 2024 threat report. This wasn’t just text generation. The operation was unusually prolific with DALL·E, pairing short comments and long-form articles with eye-catching, cartoon-style imagery designed to boost engagement on a network of fake news websites and social media profiles.

The content strategy was a geographic patchwork. English-language output from sites like britishtalks[.]com and euronewstop[.]co[.]uk mixed non-political UK cultural coverage with sharp criticism of support for Ukraine. In French, the focus shifted to West African music and culture, peppered with attacks on France’s Olympic organization—a classic Russia-origin narrative since the country’s Olympic ban. Russian-language material was more mundane, pushing commercial services in the St. Petersburg region. It’s a messy portfolio, but the anti-Ukraine and anti-France threads are the consistent, telling stitch.

What sets this operation apart is its bizarre, old-school-meets-new-school tradecraft. While the content was generated with modern models, many of the fake social media profiles used GAN-generated faces—a technique that’s been a staple of online deception since at least 2019. This anachronistic mix shows how threat actors aren’t just adopting the latest tools; they’re layering them onto dusty playbooks. The real head-scratcher, though, is how these fake outlets managed to secure “information partnerships” with real, brick-and-mortar organizations, including a school in Wales and a church in Yorkshire, which listed the arrangements on their official websites.

Despite the creative effort, the operation largely failed to capture an audience. Most of the identified X accounts had follower counts in the double digits, and Meta reported a paltry 2,100 total followers across its Facebook and Instagram assets. The low engagement is a familiar outcome for these ham-fisted campaigns, but the partnership angle introduces a more insidious form of social proof. It suggests the endgame wasn’t just about going viral, but about building a veneer of institutional legitimacy—a tactic that could pay dividends in a slower, more targeted disinformation campaign down the line.