The past week in the world of cybersecurity was unusually eventful and, at times, surreal. From hacking government alert systems to sophisticated methods of bypassing AI analysis, attackers continue to demonstrate alarming ingenuity.
Brazilian Chaos: Hackers Triggered "Invasion" Sirens
Perhaps the most striking incident occurred in Brazil. On the night of June 20, 2026, the national emergency alert system Defesa Civil Alerta was hacked. Residents of several states received highest-priority messages (Alerta Extremo) on their smartphones, accompanied by a deafening siren wail that could not be silenced even by silent mode. Instead of warnings about floods or hurricanes, the text contained incoherent phrases, slang, the word "misanthropy," and in some regions, even messages about an "alien attack." The attackers likely compromised the accounts of Civil Defense employees, gaining access to the Cell Broadcast mechanism. To stop the spam attack, authorities had to forcibly shut down servers at 1 a.m. This incident is a stark demonstration of how vulnerable critical infrastructure can be if access is not protected by multi-factor authentication and strict control.
Botnet on Abandoned Equipment: D-Link Under Attack
Canadian intelligence obtained a court order for the first time to remotely clean infected citizen devices. The target was botnets using compromised home routers and IoT gadgets to mask traffic. Researchers from XLab confirmed the seriousness of the threat by discovering a new botnet, AryStinger. It attacks outdated D-Link models (DIR-850L and DIR-818LW), turning them into proxy servers. Over 4,000 devices have already been compromised. In addition to relaying traffic, AryStinger can intercept DNS requests, monitor network activity, and steal data. Nearly half of all infections are concentrated in South Korea, China, Sweden, Malaysia, and Singapore. This is a classic example of how manufacturers, by discontinuing support for older models, leave users to face cyber threats alone.
Gaslight: New macOS Virus Deceives AI
SentinelOne specialists discovered macOS malware named Gaslight, which they believe is linked to North Korean hackers. This is not just an infostealer, but a true "gaslighter" for automated analysis systems. Hidden inside the file is a 3.5 KB loader containing 38 fabricated system messages formatted in Markdown. These messages mimic memory overflow errors, crashes, and token expiration warnings. When analyzing the sample, the AI model receives a prompt injection that causes it to doubt its own correctness and terminate the session, concluding that the file is corrupted. This is an elegant and dangerous method of bypassing AI defenses, which will undoubtedly be copied.
Europol Strikes: $47 Million and 27 Million Credentials
A coordinated operation by Europol and law enforcement from a dozen countries led to the dismantling of a network distributing the SocGholish, Amadey, and StealC malware. The results are impressive: 326 servers and 142 domains were seized, crypto assets worth over $47 million were frozen, a database with more than 27 million stolen credentials was confiscated, and approximately 15,000 compromised WordPress sites were cleaned. Of particular interest is the arrest of 69 people in Hong Kong who were involved in laundering $25.6 million through a network of shell accounts and cryptocurrency. This shows that law enforcement is increasingly intercepting the financial flows of cybercriminals.
ZachXBT Back in Action: Identity of Polish Hacker Revealed
Polish police, with support from the FBI, arrested four members of a group specializing in SIM swapping and cryptocurrency theft. The damage amounts to tens of millions of zlotys. The identities of those detained have not been officially disclosed, but renowned on-chain researcher ZachXBT identified one of them as Wojtek Kulish (nickname Merry). He made this conclusion by comparing designer items seized during the search with those Kulish had previously posted on Instagram. Another piece of evidence that hiding one's identity is becoming increasingly difficult in the digital age.
Expert Summary
The week showed that cybersecurity is no longer just a technical problem. Attacks on alert systems and the use of AI to deceive AI are already playing on the edge of psychology and technology. The only way to protect yourself is not to rely on a single layer of defense, whether it's an outdated router or an automated code analyzer. We need to think one step ahead of hackers, and as you can see, this is becoming increasingly difficult.