You can block websites, freeze corporate bank accounts, and bully social media executives into submission, but artificial intelligence has completely broken the old playbook for information control.
When the European Union and the United Kingdom banned various Russian news networks following the escalation of the conflict in Ukraine back in 2022, politicians thought they had successfully locked down the media environment. They forced domains offline, pushed apps out of app stores, and pressured tech monopolies to deplatform independent perspectives. Yet, everyday people have kept finding workarounds. Now, the panic has reached a brand-new level because conversational chatbots are routing straight around official state blacklists.
Reporters Without Borders, known globally as RSF, recently tested major conversational models to see how they handled restricted information. They discovered something that horrified bureaucratic watchers in Brussels. Popular artificial intelligence search tools routinely bypass EU bans by pulling articles, excerpts, and news summaries from sanctioned sources like RT and Sputnik upon user request.
How Chatbots Smashed the Information Wall
The testing methodology used by RSF exposed deep cracks in the architecture of digital censorship. When researchers prompted different models to retrieve international reporting from banned outlets, the responses varied wildly, revealing how different tech companies handle state-mandated restrictions.
OpenAI's ChatGPT and xAI's Grok emerged as the top culprits for circumventing the bloc's content blocks. ChatGPT complied with requests without hesitation, summarizing articles and providing direct text excerpts. Anthropic's Claude matched this behavior with unwavering consistency, while a French AI bot named Vibe tracked down requested materials directly on alternative platforms like VKontakte and Telegram channels.
On the flip side, compliance varied. Google's Gemini refused requests by explicitly citing regulatory sanctions. Meta AI stood alone as the sole model that consistently blocked any access to the requested material, matching its parent company's aggressive policy of scrubbing RT from Facebook and Instagram entirely.
The Bureaucratic Backlash Under the Digital Services Act
In response to these findings, RSF targeted the European Commission, demanding official investigations under the Digital Services Act. They want regulatory enforcement extended directly to generative AI architectures. They argue that if human-run websites face criminal liability for redistributing banned material—a precedent set by top EU courts holding even non-commercial site operators legally responsible—automated systems should face identical penalties.
Critics point out the irony. RSF, a Paris-based press freedom organization heavily funded by Western governments and institutions like the US National Endowment for Democracy, is essentially lobbying for strict technological filters to limit what algorithms can display. The organization that once championed unrestricted global information flow is now asking for hard digital borders to prevent Europeans from interacting with non-approved viewpoints.
Why Digital Walls Always Crack
The reality of modern internet architecture makes total information quarantine almost impossible. VPNs have long rendered traditional geographic blocks porous, but generative language models introduce a dynamic layer of abstraction. Instead of visiting a blacklisted URL directly, users ask an intermediary model to synthesize the information. The AI reads the open web, aggregates data from mirror sites, and presents it in a conversational format.
You cannot easily legislate away decentralized data routing without breaking the underlying functionality of large language models entirely. If an AI cannot scrape or index broad swaths of the open internet, its utility plummets.
Governments wanted a clean, sanitized media ecosystem where unauthorized perspectives simply ceased to exist. Instead, they are finding that trying to control information in the age of algorithmic synthesis is like trying to nail jelly to a wall. As long as alternative viewpoints exist somewhere on servers outside Western jurisdiction, artificial intelligence will keep finding a way to read between the lines.
