Glossary
What is personal data and AI
In short
Personal data and AI is the set of rules and risks around the way neural networks handle information about people: what you may send to a model, where to store the data and what legal basis 152-FZ, Russia’s personal data law, requires.
Security and law · personal data
MMOOZZGG.com editorsChecked: 2 min read
For a child
Every person has things only close people know: an address, a phone number, a birthday. You should not tell such things to everyone, because bad people can use them. A computer helper can also learn a lot about a person if someone types it in. So adults agreed on rules: what about people can be trusted to a computer and what must be kept safe.
In plain words
Personal data is any information that identifies a person: a name, a phone number, an address, a contract number, a voice recording. When a company connects a neural network to customer conversations or to a staff database, this information starts to pass through the model. Russian law requires a basis for the processing, such as the person’s consent, and requires that data of Russian citizens be collected and stored in databases inside the country. If you send such conversations to a foreign service, you can break the law even if no leak happened.
For business
This issue is the most common reason a security team stops an AI project. They check three things. Where the data is physically processed, and whether Roskomnadzor must be notified before a cross-border transfer starts. Whether there is a basis for processing: consent or a contract with the customer. Whether the system makes decisions about people by itself: decisions with legal consequences based only on automated processing are allowed only with written consent or a direct basis in law. The protection options are a model in a data center in Russia or inside your own network, anonymization before sending, and a data processing agreement with the contractor. Settle these questions before the project starts, not after the first audit.
For a specialist
Federal Law No. 152-FZ of 27 July 2006 “On Personal Data” applies to AI systems in the same way as to any other processing. Part 5 of Article 18 forbids recording, organizing, accumulating and storing personal data of Russian citizens in databases outside the country during collection, except in certain cases. Article 12 requires notifying the authorized body before a cross-border transfer starts. Article 16 forbids decisions with legal consequences based only on automated processing without written consent or a basis in federal law, and requires giving the person a way to object. The grounds for processing are listed in Article 6: consent, performance of a contract with the data subject and others. For LLM systems, this becomes a set of design choices: inference and logs in Russian jurisdiction, a minimum of fields in the prompt, pseudonymization or masking before an external API, access control in RAG, retention periods for dialogs and no training of the vendor’s model on this data.
Example
A clinic wants an AI agent to confirm patient appointments. The agent runs on a server in Russia, and instead of the full name and phone number, the model receives placeholder labels that are replaced with the real values back in the booking system.
Related terms
Questions and answers
Sources
- Federal Law No. 152-FZ of 27 July 2006 "On Personal Data", Article 18, ConsultantPlusconsultant.ru
- Federal Law No. 152-FZ, Article 12 "Cross-border transfer of personal data", ConsultantPlusconsultant.ru
- Federal Law No. 152-FZ, Article 16, ConsultantPlusconsultant.ru
- Federal Law No. 152-FZ, Article 6 "Conditions for processing personal data", ConsultantPlusconsultant.ru
Not sure whether your company needs this?
AI solutions map