Column by Tatyana Isakova about GPTBot

This November, the traditional month of sales and increased load, including malicious, on marketplaces, cybersecurity specialists discovered an unusual “visitor” in the networks of large Russian companies. According to Qrator Labs, the most common bot visiting the websites of large companies during this period was the OpenAI bot – GPTBot, which showed such activity for the first time in Russia.

GPTBot was launched in August to crawl the web and collect open data, which the American company will later use to improve its acclaimed ChatGPT neural network. And although OpenAI closed developments for Russia (the company’s website is unavailable in the country), as it turned out, it did not bypass the domestic segment of the Internet. “For a number of large online stores, the share of GPTBot requests in the mass of all bot requests in November reached 90%,” said Qrator Labs.

Abroad, GPTBot’s curiosity was encountered a little earlier, which led to criticism of the product and the company for unauthorized data collection: organizations and users are afraid of training a neural network on their content. Even OpenAI’s public announcement that the site could be closed to GPTBot and the attached instructions on how to do this did not help. In Russia, the bot harmed companies rather through an increase in the already high load on servers and the costs of maintaining the functionality of sites. One of the marketplaces complained to me about this, although neither Ozon, nor Wildberries, nor Yandex Market officially commented on the situation.

OpenAI’s interest in Russia can be explained by its reluctance to miss out on the Russian-language segment of the Internet, my interlocutors in the IT market are sure. Even if the company does not lift the blocking of its technologies for the Russian Federation, its algorithms will be able to be used in other countries where the Russian-speaking audience, for obvious reasons, has increased over the past year and a half.

Using Russian OpenAI resources can be convenient and profitable. Class action lawsuits against the company from famous writers such as George R.R. Martin and John Grisham have already caused a stir. Russian organizations and users have much less chance of even filing such a complaint, not to mention winning the dispute. Therefore, familiar IT specialists assure that domestic open data is steadily becoming part of the global dataset.

In the future, this may negate, among other things, the efforts of the authorities to limit the collection of data by Western corporations – this is precisely what the authors of the landing law and supporters of blocking foreign social networks fought for because of their recommendation algorithms. As it turned out, there were completely different products to be wary of.