Monitoring law changes
Compliance starts with the question of what has changed. Every day, AI reads official publications of legal acts and drafts, picks those that concern the company’s industry and processes and writes a short note: what changed, from what date and which internal documents it affects.
An example from 2026. Federal Law No. 243-FZ of 26 July 2026 “On Supporting the Development of Artificial Intelligence Technologies in the Russian Federation” took effect on 1 September 2026. Under it, data of sovereign and national models is processed in Russia, and the Government decides the cases where such models are mandatory.
For compliance, this creates two tasks. First, find out which AI services employees already use and where their requests are processed. Second, follow the Government acts that will name the cases of mandatory use.
What a business should do about the law on AI is covered in a blog article, and the overall picture is in a review of the legal regulation of AI for lawyers.
The note uses only official sources: the legal information portal and the websites of government agencies. AI treats news retellings as a signal, but the note links to the text of the act itself and its version.
A compliance officer reads the AI note. The officer decides whether the change affects the company and names the person responsible for editing the document.
Checking policies
Internal policies go out of date quietly: the law changes, and the personal data regulation stays in its old version for years. AI compares each policy with the current rules and flags gaps with the article number.
A typical check under 152-FZ, Russia’s personal data law, has four questions:
1. Does the policy describe the procedure for cross-border transfer? Since 1 March 2023, an operator must notify Roskomnadzor, the Russian data protection regulator, before sending personal data abroad (part 3 of article 12 of 152-FZ).
2. Where is the data of Russian citizens recorded and stored? Under part 5 of article 18 of 152-FZ, collection and storage take place in databases inside Russia.
3. Which outside services, AI services included, receive employee and customer data, and is there a data processing agreement with them?
4. Who is responsible for an incident, and how is it reported inside the company?
The cost of a gap is measurable: the fine for a company under part 8 of article 13.11 of the Russian Code of Administrative Offenses reaches $72,000. AI prepares a table “policy clause, rule, gap, proposed edit”, and a lawyer approves the new version.
Employees need a separate document, an AI use policy: which services they may use and what they must not send to them.
Prohibited terms in contracts
A prohibited term is a clause the company does not sign under its own rules. For example, unlimited liability, data transfer to a contractor without a processing agreement, data storage abroad, or termination by the counterparty without compensation.
The legal team keeps the list of such terms, and AI applies it to every contract. The check covers both incoming drafts and the archive of current contracts.
For the archive, the result looks like a register: the contract number, the clause, the rule it breaks and the date when the contract can be changed or ended.
A separate line covers contractors that receive personal data. Since 30 May 2025, under 420-FZ, a company pays 1% to 3% of revenue for a repeat leak, but at least $240,000 and at most $6 million (part 15 of article 13.11 of the Code of Administrative Offenses).
So a contract with a contractor states the place of processing, the security measures and the incident notice procedure. If these terms are missing, AI puts the contract at the top of the list for renegotiation.
AI does not edit the contract itself. It marks the clause, suggests wording from the company’s approved templates and passes it to the lawyer who handles the contract.
Report for management
Management needs a risk register, not a list of articles: what is broken, what it may cost, who fixes it and by what date. AI builds it from the results of the three checks above.
Sample register lines:
1. The personal data processing policy does not describe cross-border transfer. Risk: a fine of up to $72,000 under part 8 of article 13.11 of the Code of Administrative Offenses. Owner: the legal team.
2. Contracts with contractors that receive personal data have no leak notice term. Risk: a revenue-based fine under 420-FZ for a repeat leak.
3. Employees use a foreign AI service, and the place of data processing has not been assessed. Risk: a breach of part 5 of article 18 of 152-FZ. Owners: IT and compliance.
4. Under 243-FZ, the company awaits Government acts on the cases of mandatory use of sovereign and national models. Action: a list of the systems this may affect.
The register updates with each change in the law and each new contract, and the compliance officer approves it before it goes to management. People decide on the measures, and AI prepares the basis.
How AI fits into the whole work of the legal team, from contracts to claims, is covered on a separate page.