FAS: from “Big Digital Cat” to “Anticartel”
FAS launched the first module of its “Big Digital Cat” system in the fall of 2019. The service announced the development in 2018, and “Anticartel” was described from the start as an upgraded version of the same system.
The Anticartel state information system has worked since 1 August 2025. It takes data from the Unified Procurement Information System and the state Auctions system and, according to FAS, analyzes 100% of procurement procedures every day with elements of machine learning.
The system’s task, as the agency describes it, is to prevent, detect and stop cartels with digital technology and AI, and also to cut the number of requests the service sends to market participants.
FAS described the machine’s role back in 2024. The second stage of the project includes analytics with machine learning and AI but no automatic decisions on violations: the service treats automatic conclusions as auxiliary, and commissions consider the cases.
The agency named the result too. According to the deputy head of FAS, the share of cartels in tenders uncovered with information from the system was 21% in 2025 and reached almost 40% by May 2026. Now the system is being integrated with the systems of other state bodies.
Which signs of collusion show up in tender data and how they are checked is covered on the page about bid rigging.
The Audit Chamber: digital audit
On 12 December 2025, the Chairman of the Audit Chamber said that in 2026 the agency plans to launch a state information system for digital audit. According to the press service, the chamber’s main work with AI centers on this system, the unified digital platform “Digital Audit”.
In 2026, the chamber is adapting AI for risk-based audit, and by 2030 it plans to build it into its core processes. Risk-based audit means checking first those whose data points to a risk, not everyone in a row.
Regional audit bodies will get access to the system too. Through personal accounts, they will be able to analyze data and use state audit methods without developing their own tools.
How individual AI agents help check acts, contracts and payments is described on the page about AI in the work of audit bodies.
The treasury and document checks
The second part of spending control is documents: acts, invoices, payment orders. Here AI works as a fast reader that compares the papers with each other and flags discrepancies.
At the Audit Chamber, according to its deputy chairman, the system reads acts and payment documents, structures the data and saves tens of thousands of working hours (June 2026).
In Moscow, a “Digital assistant” for the treasury has worked since 2024. According to the Moscow Government, it checks financial documents, finds inconsistencies and prepares protocols for payment decisions; document processing accuracy is above 90%.
A check before payment changes the order of control: an error or a forged document is found while the money has not yet left. An employee makes the payment decision, and the assistant prepares a protocol for them.
Such systems do not assess people. They compare the details, amounts and dates in documents and show where they do not match, and a specialist has to sort out each discrepancy.
Kazakhstan: prices and choice of supplier
The Anti-Corruption Service of Kazakhstan, together with the Atameken chamber of entrepreneurs, watches procurement prices through the “Single Procurement Window” system. According to the service, in two years the monitoring of inflated prices prevented wasteful spending of more than KZT 67 billion: KZT 29 billion in 2022 and KZT 38.4 billion in the first 10 months of 2023.
The sign here is clear: the planned price is far above the market price. In one of the service’s examples, a hospital planned to buy 16 portable medical lamps at KZT 3.3 million each, while the market price was up to KZT 600,000. The purchase was canceled, and the savings were KZT 44 million.
Other examples of the same type from the service: a yurt for regional sports events, an air conditioner for a district house of culture, and diesel fuel for schools and kindergartens. In all cases, the planned price was well above the market price, and for the yurt and the air conditioner the budget was revised.
A price above the market is not yet a violation: there may be a special configuration or an urgent delivery. So the price signal goes to the buyer and the inspectors, and they decide whether to revise the purchase or cancel it.
The Ministry of Finance of Kazakhstan selects suppliers with an AI rating system linked to 28 databases. According to the vice minister of finance, the system assesses participants’ experience, taxes paid, presence in registers and place of registration, and it forms the final decision without a person.
This approach differs from that of FAS, where the machine only gives hints. Russian state bodies can use both parts of Kazakhstan’s experience: comparing the price with the market as a signal for a check, and collecting supplier data from many registers.
What else can be transferred from global practice
The systems that Russian agencies talk about openly focus on procurement and documents. The data for the next steps already exists: the same Unified Procurement Information System and state Auctions system that FAS reviews every day, plus the Unified State Register of Legal Entities and the registers of support recipients.
1. Checking programs before launch. In the United Kingdom, an AI tool looks for weak points in drafts of new government programs; in Russia, the same approach could review subsidy rules before the first payment.
2. Checking payments against registers. Payments to dead or non-existent recipients are found by comparison with the civil registry and HR data, as was done in Colombia and Nigeria.
3. A graph of links. Shared phone numbers, representatives and addresses of companies are collected into one model to see who is linked to whom in tenders and in the chain of payments.
Examples that can be transferred to Russian practice, with an analysis of the conditions, are on the page about transferring global practice. The overall picture of AI in public administration is in a separate review.