The expert gave advice on how to implement AI in a company without the risk of data leakage
- Новости
- Society
- The expert gave advice on how to implement AI in a company without the risk of data leakage
Artificial intelligence (AI) is becoming a part of companies' workflows, helping them prepare documents, analyze data, and automate routine tasks. However, the use of neural networks can lead to the leakage of confidential information if employees transfer contracts, customer bases, or internal documents to external services. Sergey Taran, Managing Director of LANSOFT, told Izvestia about this on September 10.
According to the expert, the main risk is not related to the use of AI itself, but to the lack of clear rules for working with data. An employee can upload a contract fragment, technical documentation, or financial statements to the neural network, after which the company loses control over where this information is stored and who can access it.
"The first stage of AI implementation should be data inventory. The company needs to divide the information into categories: public, internal, and restricted information, including personal data and trade secrets. Each group should have its own rules. Open source materials can be used to prepare drafts, search for ideas and process information, whereas internal documents should work only in a secure corporate environment," Taran said.
According to the expert, personal data of clients and employees, transaction information and critical information should not be transferred to open external services without appropriate legal and technical grounds.
However, a simple ban on the use of neural networks does not solve the problem. According to Taran, employees can still access public tools from personal devices. A more effective approach is to create an enterprise AI service with access control, an action log, and settings that exclude the use of internal data to train third-party models.
The expert noted that companies can also use open artificial intelligence models that are installed inside their own infrastructure. Such solutions allow you to maintain control over information, but require additional technical resources.
Another important principle of safe work with AI is data minimization. According to Taran, the algorithm does not always need a complete document or database to complete the task. For example, when analyzing a client's request, you can delete the last name, phone number, address, and other personal information. In documents, you can replace real data with symbols to get the desired result without the risk of disclosing confidential information.
The expert paid special attention to setting up access rights. According to him, employees and AI systems should receive only those powers that are necessary to perform specific tasks. For example, the algorithm can prepare a draft response or generate a report on a specific folder, but it should not be able to download the entire customer base on its own or change critical data without additional confirmation.
Employee training remains equally important. According to Taran, general prohibitions like "do not upload sensitive data" work worse than clear instructions. Companies need to explain which services are allowed, what data can be used in requests to neural networks, how to properly depersonalize information and where to contact in case of errors.
The expert emphasized that the safe use of AI requires monitoring not only incoming requests, but also the results of the algorithms. It is important to identify unreliable answers, limit the use of AI in sensitive areas and prevent its uncontrolled impact on information systems.
According to Taran, the effective implementation of artificial intelligence is not based on a complete ban on technology, but on a combination of clear rules, secure infrastructure and employee training. This approach allows you to take advantage of AI while maintaining control over corporate information.
Anastasia Gorelkina, Chairman of the Board of Directors of JSC HC Siberian Business Union, member of the Board of Directors of JSC Azot, recalled on August 25 that the value of an employee for a company is determined not only by the position and salary level, but also by a combination of rare competencies, influence on important business processes and the complexity of his replacement. According to the expert, today the labor market is particularly in demand for specialists in the fields of data and AI, cybersecurity, automation, product development, complex project management and sales with access to key customers.
Переведено сервисом «Яндекс Переводчик»