Babushkinsky District Court of Moscow has recognized the dismissal of Ekaterina Remizova, the sales director of the company "GK Energoprof," for disclosing trade secrets as lawful. The employee was forwarding official documents to her personal email and uploading them to the DeepSeek neural network. This is reported by RAPSI and RIA Novosti.
Remizova worked at the company from April 18, 2025, with a salary of over 800,000 rubles per month. In September 2025, she was already subject to disciplinary action, which was followed by dismissal for disclosing a secret protected by law.
What the Employee Was Charged With
The employer established that Remizova was sending documents regarding participation in a tender and a file with colleagues' salary data to her personal email address. She also used her personal address as a hidden copy (BCC) when forwarding corporate correspondence. In addition, she was extracting protected information from the company's internal resource—including PowerBI reports with statistics on deals, incoming funds, and financial indicators of departments—to the third-party service DeepSeek, developed by the Chinese company of the same name.
The court emphasized that the mere act of uploading trade secrets or confidential information into the Deepseek AI system already constitutes disclosure of such information.
An additional reason for dismissal was the disclosure of the company's strategic plans to the supplier "Sibtorgservis": Remizova informed them of the intention to launch the "Energogranum" brand on the market and establish direct supplies of generators from China, after which the counterparty ceased communication. The employer also cited systematic failure to meet sales targets.
The Demand for a "Golden Parachute" and the Outcome of the Proceedings
After her dismissal, Remizova filed a lawsuit demanding that the order be declared unlawful, the grounds for dismissal be changed to "by mutual agreement," and the "golden parachute" provided for in an additional agreement, amounting to 5 million rubles, be recovered. The company did not admit the claim. During the proceedings, the defendant offered a settlement agreement with a payment of over 400,000 rubles—half of the salary—but the plaintiff refused it.
The court stated that Remizova had no production necessity to extract information from the employer's protected media, send it to an uncontrolled external address, or post it on the third-party resource DeepSeek. These actions were classified as a gross violation of labor duties.
Outcomes of the case:
- The demand for payment of the 5 million ruble "golden parachute" was rejected, as this payment is provided only upon termination of the contract by mutual agreement or staff reduction;
- One of the disciplinary penalties was deemed unfounded; the company was ordered to pay 10,000 rubles in moral damages compensation;
- The remainder of the claim, including the request to declare the dismissal unlawful, was denied.
The court's decision on case No. 02-1545/2026 has not yet entered into legal force and can be appealed.
AI Opinion
From the perspective of machine data analysis, the Remizova case reflects only a part of a broader problem: the DeepSeek service allows data leakage not only through closed file uploads but also through its own Share function. Information security specialist David Konitzny discovered that the Google search engine freely indexes dialogues from the directory chat.deepseek.com/share/, and they can be found with a simple query: site:chat.deepseek.com/share/. The system's warning that the link can be opened by "anyone who has it" says nothing about possible indexing by search engines—and it is in this gap between the expectation of privacy and the technical design of the service that an additional risk lies for companies whose employees use such tools.
It turns out that even without malicious intent, a temporary link can turn into a permanent digital trail with fragments of code, work tasks, and financial data. How common is the practice of accidental disclosure of corporate information through such "temporary" links in other neural network services?
