"Astra Group" presented a set of solutions with which a company can launch neural networks not on external servers, but on its own – and link them with its data and documents. The company announced this at the press conference "Course for AI Platform".
The artificial intelligence market in Russia grew by 31% in 2025, reaching 316.1 billion rubles, while generative services grew almost fivefold, to 58 billion. But few benefit: 97% of companies try AI, 10-15% bring it to permanent operation, and more than 90% do not get a return. The reason is that 79% of critical infrastructure enterprises prohibit employees from using public chatbots. People still use them and pay out of their own pocket – this is called "shadow AI": data goes out, there is no budget or control. "Astra" proposes placing models inside the company's protected perimeter and using various ones: the most powerful, proven open-source, and Russian.
Technical Director Anton Shmakov reduced security to simple rules: data does not leave the company, employees have limits and access rights, leaks are monitored, and what the model outputs is checked. The Astra Linux operating system, on which everything runs, isolates AI assistants from each other and limits their actions. It is used by more than 30,000 customers.
"Astra" has already transitioned its teams to AI: in one department, labor intensity decreased by half, productivity increased by 70-90%, in others – by 20-30%. But the effect depends on the type of project: for new developments, the gain is 35-40%, for old code – less than 10%. According to the company's calculations, if AI is implemented selectively, the increase will be about 5%, if throughout the entire development cycle – about 50%, and costs will pay off in about six months.
Read more on the topic:
- Russian AI market soared by almost 30%: it doubled in two years, reaching 316 billion rubles
- Astra Linux Creator Invested 100 Million Rubles in Neural Networks: The Company Launched a New AI Direction
- Russian Business Rolled Back 9 out of 10 AI Implementation Projects - Survey Data