French researchers were convinced of this when they tested the Russian version of GPT Chat, Ukrinform reported with reference to the TF1 TV channel.

“The blocking of everything related to Ukraine goes so far that even old Russian proverbs are being censored because they refer to places in Ukraine,” Kevin Limonnier, senior lecturer in Slavic studies at the French Institute of Geopolitics, told reporters.

“It’s just kidding and it’s definitely much better than ChatGPT,” Yandex’s French AI testers say, quoting Russian user comments. The tool, called “YAGPT2,” can also use artificial intelligence to lead discussions and answer users’ questions, its developers said. But once the conversation turns to political topics, the conversation becomes unresponsive, according to researchers. He also doesn’t know where some cities are in Ukraine, can’t say who Zelenskyi is, and avoids asking questions about Oleksiy Navalny.

In response to the question of who the current president of Ukraine is, the artificial intelligence says that it “understands that the answer to this question would be very useful”, but “does not discuss such issues in order not to offend anyone.” The same thing happens when the chatter is asked to summarize the events after the annexation of Crimea by Russia in 2014. Or when they ask about the beginning of the “war in Ukraine”.

Yandex warned on its website that its broker is “unethical” and will not respond to certain requests. However, according to French researchers, such demands are surprisingly high. The conversation tells, for example, who is the president of Russia or France, but excludes the possibility of explaining diplomatic relations between the two states in a few words.

Several dozen questions showed that the system avoids discussing all issues related to Ukraine and the West. Regarding domestic policy, the chat is censored so that he can mention the name of the country’s agriculture minister, but he refrains from naming the defense minister.

The blocking is such that the tool eliminates even simple geographical questions. He specifically refused to give Avdiivka’s location, claiming he “didn’t want to offend anyone.”

Reportedly, in the Netherlands, artificial intelligence is answering questions from patients of one of the hospitals there, they say that with its help, the consultation process is significantly accelerated and more time appears for offline patients.