The Russian Federal Tax Service (FNS) is increasingly scrutinizing IT companies by analyzing their actual business operations rather than just formal documentation. Inspectors examine the origin of software, the physical location of developers, and the consistency between legal contracts and technical data from platforms like Git, Jira, and Confluence.
A major focus for the FNS is the structure of corporate groups. The tax authority investigates whether splitting a business into multiple entities is driven by legitimate goals or is merely a tactic to reduce tax burdens. The case of Demis Group highlights that courts evaluate the cumulative effect of business signs, such as shared management, common resources, and identical client handling.
Companies often face risks when legal structures do not match operational reality. Common red flags include shared bank access for multiple entities, transferring employees to independent contractor status while keeping them in the same roles, and centralized management of subsidiaries. The FNS assesses whether a company would maintain its current structure if tax rates were equalized.
The source also mentions a comparison of AI models based on the 'Cost per Task' metric. According to the data, the GPT-5.6 Luna (max) model costs $0.18 per task, while the GPT-6 Astra (max) model costs $3.26. The Artificial Analysis Intelligence Index for these models is 38 and 53, respectively, indicating a significant difference in efficiency and cost for developers.
