How Rosatom Bought AI Tools Through Unofficial Channels

The author of a Habr article tells the story of how, judging by the headline "How Rosatom Bought AI Tools Through Unofficial Channels," some of those tools at Rosatom were purchased outside standard procurement procedures. The text does not open with a dry account of facts but with a series of proverbs in different languages about the same phenomenon: cutting corners on important things ends in far greater losses. In Russian it is "кроилово ведёт к попадалову" (roughly, "skimping leads to a costly fall"), in Polish "Tanie mięso psy jedzą" ("dogs eat cheap meat"), and in the British tradition "Penny wise, pound foolish."

What the author warns about at the start

• The article's headline is "How Rosatom Bought AI Tools Through Unofficial Channels"
• Source: the Habr AI section
• Proverbs in three languages: Russian, Polish and English versions of the same idea about false economy
• Key quote: the Latin question "Quis custodiet ipsos custodes?" ("who will guard the guards themselves?")
• Stated theme: saving on SMM staff as the starting point of the story

The author then cites the classic Latin question "Quis custodiet ipsos custodes?" and ties it to the savings on SMM staff which, according to the publication, set the later events in motion. This rhetorical device frames the piece: it is not merely about a breach of internal procurement procedure, but about a situation where the lack of internal control let a practice carried out outside official channels go unnoticed until its consequences became obvious to a wider audience, including readers of specialist media.

Why are unofficial purchases of AI tools a systemic risk?

The broader problem raised by the headline goes beyond this one case. In large regulated organizations, such as state corporations, banks and industrial holdings, procurement of software and AI service subscriptions is centralized precisely so that the vendor is vetted, and data-handling practices, license legitimacy and security requirements are confirmed before a tool gets access to corporate accounts or information. When individual departments or contractors bypass this process, often to save time or money, for example by using unofficial resellers or personal accounts for marketing and SMM tools, the organization loses visibility into what has gained access to its systems and cannot confirm that the tool meets security and compliance standards.

This class of problem, commonly called "shadow IT" in corporate practice, has worsened as AI tools appear and spread faster than formal approval procedures can vet them. The temptation to skip a long chain of approvals for a quick result grows with the number of AI services on the market.

What this means for large organizations

Stories like the one the author describes tend to become case studies in why governing the purchase and use of AI tools matters in the AI era just as much as it has always mattered for traditional software. That governance means vetting vendors, defining clearly who may approve a purchase, and tracking which tools have access to which data. For organizations handling sensitive or regulated information, the rhetorical question "who will guard the guards themselves," with which the author opens the article, reads not as an abstract quotation but as a direct pointer to the absence of internal control over who makes decisions about buying new digital tools locally, and how.

The choice of a trilingual set of proverbs, Russian, Polish and English, also serves the overall message: the author shows that the phenomenon is not unique to one country or industry but is a universal management trap that organizations worldwide fall into when they decide to economize on control procedures for short-term gain. That is why the headline presents the article as an analysis of a specific case, while in practice it raises a question relevant to any large body that buys digital tools outside a single center of responsibility.