A weekly roundup of the most high-profile incidents and threats in digital security. This week — from large-scale financial crimes to new attack vectors targeting mobile devices.
Brazilian drug cartel and crypto brokers: laundering billions
Brazil's Federal Police dismantled an international drug syndicate that, according to investigators, organized the smuggling of at least 6.5 tons of cocaine. A key link in their money laundering scheme was illegal financial brokers who facilitated the conversion of fiat funds into cryptocurrency and their subsequent transfer abroad.
During a large-scale operation across four states, 44 searches were conducted, nine people were detained, and another 13 remain in custody. Law enforcement froze assets worth up to 1 billion Brazilian reais (about $197 million). Notably, this operation followed recent U.S. Treasury sanctions against Ethereum addresses linked to the Sinaloa cartel, confirming the growing trend of using digital assets in transnational crime.
Cyberattack on Minnesota water utilities: a strike on critical infrastructure
On July 26–27, unknown hackers carried out a targeted attack on the operational technology of more than 30 public water systems in Minnesota. The incident forced the local IT agency (MNIT) to activate cybersecurity protocols at the state level. In the city of Braham, the water treatment plant came to a complete halt for three hours, while in other counties, staff had to switch pumps to manual control.
Authorities emphasize that water treatment quality was not affected. The investigation is being conducted jointly with the FBI and the U.S. Cybersecurity and Infrastructure Security Agency. Earlier, U.S. agencies had warned of increased activity by Iranian hackers targeting Rockwell Automation programmable logic controllers, indicating the systemic nature of the threat to critical infrastructure.
Ukrainian "crypto academy": fraud totaling $1.11 million
Ukraine's cyber police exposed an organized group operating under the guise of the "Ukrainian Financial Academy." The perpetrators deceived about a thousand people, stealing $1.11 million under the pretext of providing training in cryptocurrency investments. Four suspects, including two organizers, were arrested after searches in the Kharkiv and Dnipropetrovsk regions.
The scheme was classic but well-oiled: mass advertising on social media, a fake trading platform with "successful" charts, and then blocking withdrawals under the guise of fees. What made it particularly cynical was that the fraudsters deliberately targeted the most vulnerable victims — pensioners and the seriously ill. One victim, undergoing cancer treatment, transferred about 500,000 hryvnias that she had saved for therapy. The suspects face up to 12 years in prison.
SparkKitty: a new stealer hunting seed phrases in the gallery
Check Point specialists discovered the SparkKitty malware, which infiltrated devices through official app stores. Instead of intercepting keystrokes, this stealer requests access to the smartphone's gallery and uses an OCR engine to scan photos for crypto wallet seed phrases, passwords, and QR codes.
On iOS, the virus was found in the 币coin app, and on Android — in the SOEX messenger, which was downloaded more than 10,000 times before being removed from Google Play. The malware also spread through third-party catalogs and modified APKs. This once again confirms a critical rule: storing wallet backups in digital form is an extremely risky practice. The only safe method is offline storage.
DeepSeek AI agent in autonomous cyberattacks
Unit 42 researchers uncovered a campaign in which a Chinese-speaking hacker used the DeepSeek AI model to conduct fully autonomous attacks. Controlling the neural network via Telegram, the operator only entered commands, while the AI independently scanned the internet through the FOFA search engine, sought out vulnerabilities, and applied exploits from GitHub.
The autonomous attempts did not fully succeed, but in manual mode, the hacker extracted data from three organizations via a NetScaler vulnerability and compromised 11 Marimo targets. In total, more than 460 targets were attacked. The attacker's infrastructure was exposed due to his own mistake — an accidentally launched open HTTP server. This case demonstrates that AI agents are becoming a real tool in cybercrime, significantly lowering the barrier to entry for conducting sophisticated attacks.
My comment: The trend toward using AI in autonomous attacks and integrating cryptocurrencies into traditional criminal schemes requires the industry to rethink its approach to security. We are moving toward a world where protection must be no less adaptive than the threats — and this applies both to institutional investors and ordinary users.