Russian Federal Tax Service Tests Scoring Model to Uncover Concealed Self-Employed Income
The Federal Tax Service of Russia is testing a new scoring model aimed at identifying self-employed individuals who conceal income by unlawfully canceling receipts. Deleting a check in the Moy Nalog app does not remove the transaction from the tax authority’s view.
According to Vladimir Maltsev, head of the operational control department, the system evaluates multiple indicators. Systematic cancellations over a defined period represent one key signal. Another trigger occurs when a partner—such as a digital platform or bank—annuls a receipt issued by the self-employed person.
The complete set of parameters and their respective weights have not been disclosed. The model is currently operating in pilot mode across selected regions.
During the first nine months of 2026, roughly 3,000 self-employed taxpayers were held accountable for improper receipt cancellations, with total penalties reaching approximately 300 million rubles. Officials note that regularly nullifying recorded revenue is becoming a less viable strategy.
The initiative applies risk-scoring methods previously common in banking directly to receipt data. Tax authorities now examine not only the volume of incoming payments but also how users handle documented operations.
Related articles
Reviving Huawei SANEX9510: Expired Certificates, Zero-Byte Keystore and Lost '@' Symbol Fixed Without Database Restart
A three-day postmortem revealed seven independent failures across four TLS contours in a legacy Huawei SANEX9510 telecom OSS platform that had been offline for ten months. The issues included expired certificates, an empty 0-byte PKCS#12 keystore, a missing '@' character in an encrypted TrustStore password, incompatible PKCS#8 private key encryption formats, and connection leaks to Sybase IQ. All components were restored without rebooting servers, restarting the database, or resetting passwords. The engineer mapped every keystore reference across multiple property files to avoid breaking working segments while fixing others. A final table creation for missing DAS_DIRSIZE metadata resolved a month-old import failure unrelated to the certificate problems.
Courtroom Sketches in the US and UK: Artistic Traditions, Charles Manson Trial and Lawyer Irvin Kanarek
The article explores the history and techniques of courtroom sketching in the United States and Britain, highlighting key artists and their distinctive styles. It details how Bill Robles captured the dramatic moment when Charles Manson lunged at Judge Older during his trial. The piece also reveals the unusual background of Manson's third attorney, Irvin Kanarek, a former rocket engineer turned lawyer known for deliberately prolonging cases. British artist Priscilla Coleman is noted for her method of sketching from memory after taking detailed notes in court. Additional artists such as Jane Rosenberg, Arnold Mesches, and Joseph Papin are profiled for their contributions to the genre. The text emphasizes that courtroom art conveys emotion and atmosphere that photographs cannot capture.
No News Content Available
The provided news item contains no text or details in the original Chinese source. Without any substantive information, it is impossible to determine relevance to cybersecurity topics. The entry is therefore classified outside all security-related categories.
Kimi K3 Open Model Surpasses Closed GPT and Anthropic Systems on Independent Benchmarks
Kimi K3 is an open-weight model developed by Moonshot AI that achieves state-of-the-art results on several independent benchmarks while allowing full local deployment. The architecture scales context length to one million tokens, increases layer depth, and expands the expert count to 896 within a Mixture-of-Experts design. Developers replaced the standard KV cache with a fixed-size state matrix updated via the delta-rule and introduced stabilizing mechanisms to prevent numerical overflow in BF16 precision. Sequential state updates were parallelized across chip, card, and multi-card levels using the open-sourced FlashKDA kernel built on CUTLASS. Information flow between layers was restructured with learned pseudo-queries that selectively retrieve relevant outputs from prior blocks, while expert routing employs per-expert quantile thresholds and a latent-space projection to maintain load balance. The resulting model demonstrates that carefully engineered scaling across context, depth, and width can close the gap with closed-source leaders without requiring proprietary data or compute advantages.