Broken Webs of Knowledge

Among the myriad negative reactions to the recent release of the Chinese AI model DeepSeek, one of the most common involved what was an obvious “gotcha” query: questioning the chatbot about the 1989 student protests Tiananmen Square, and noting that – tut tut – the results were so clearly censored.
Rarely was this observation followed up with any sort of concern about how readily this could happen – could already be happening – in American AI technology. Political oppression, so the story goes, is something that happens ”over there,” with Chinese technologies under the Chinese regime; US computing, to the contrary, is engineered to express and unleash intellectual freedom.
This is, of course, demonstrably false.
The Guardian published an investigation this week into Google’s cooperation with autocratic regimes around the world in facilitating their censorship requests – the removal YouTube videos and links to websites at the behest of Russian, Chinese, Iranian, and Taliban authorities, for starters. “Google’s own data reveals that, globally, there are 5.6 million items of content it has ‘named for removal’ after government requests,” Sian Boyle reports. “Worldwide requests to Google for content removals have more than doubled since 2020, according to cybersecurity company Surfshark.”
The Internet giant was also one of the first tech companies this week to comply with President Trump’s executive order renaming the Gulf of Mexico – Google Maps now reads “Gulf of America.” And while Google found the time to update its maps product, the company also announced it no longer has the resources to maintain holidays like Black History Month in Google Calendar.