SecurityWeek’s cybersecurity news roundup provides a concise compilation of noteworthy stories that might have slipped under the radar.
We provide a valuable summary of stories that may not warrant an entire article, but are nonetheless important for a comprehensive understanding of the cybersecurity landscape.
Each week, we curate and present a collection of noteworthy developments, ranging from the latest vulnerability discoveries and emerging attack techniques to significant policy changes and industry reports.
Here are this week’s stories:
Russian man pleads guilty over Yanluowang ransomware attacks
Aleksei Olegovich Volkov, a 25-year-old Russian national, has pleaded guilty to charges related to his role as an initial access broker for the Yanluowang ransomware group in 2021 and 2022, CyberScoop reported. Prosecutors said two of the group’s victims paid a total of $1.5 million in ransoms. Volkov was arrested in Italy in 2024 and extradited to the United States, where he faces up to 53 years in prison.
Asahi brewer’s supply crippled by ransomware one month after attack
Japan’s largest brewer, Asahi, continues to suffer severe disruption to its domestic order and logistics systems more than a month after a ransomware attack by the Qilin group. The incident forced the company to revert to manual processing, cutting beer shipments to approximately 10% of regular volumes during Japan’s peak season, The Japan Times reported. The prolonged disruption has allowed competitors to gain market share.
Synology patches vulnerability disclosed at Pwn2Own
Synology released a patch for a critical remote code execution vulnerability in its BeeStation OS that was successfully demonstrated at the Pwn2Own Ireland 2025 competition. The flaw is identified as CVE-2025-12686. Researchers from Synacktiv were awarded $40,000 for discovering and exploiting the issue. QNAP has also released patches for flaws disclosed at Pwn2Own.
Amazon starts private AI bug bounty program
Amazon has launched a new private AI bug bounty program to strengthen its foundation models, including Amazon Nova. The invite-only program aims to engage security researchers and university experts to find and fix security vulnerabilities, biases, and potential for harmful activities like prompt injection and CBRN (Chemical, Biological, Radiological, and Nuclear) threat assistance. This initiative complements Amazon’s existing public bug bounty program and offers rewards ranging from $200 to $25,000.
Windows Kerberos delegation flaw allows full domain control
Silverfort discovered a new Windows Kerberos delegation vulnerability, tracked as CVE-2025-60704 and dubbed ‘CheckSum’, which affects any organization using Active Directory with delegation enabled. The flaw allows an attacker who has gained initial access to an environment to impersonate arbitrary users, escalate privileges, and ultimately gain control over the entire domain. Microsoft has issued an update as part of Patch Tuesday to address the vulnerability, which carries a CVSS score of 7.5.
Researchers uncover Sora 2 system prompt
Researchers from Mindgard successfully extracted the hidden system prompt (the core internal instructions) from OpenAI’s Sora 2 video generation model. The team accomplished this using a technique that involved asking the model to reveal its hidden instructions through text, image, video, and audio generation. While text and image-based attacks produced only fragments, audio generation (especially with transcripts enabled) allowed the researchers to stitch together a nearly complete system prompt.
Deepwatch lays off staff to boost AI
Cybersecurity firm Deepwatch has laid off between 60 and 80 employees, representing roughly a quarter of its total workforce, TechCrunch reported. CEO John DiLullo stated the restructuring is necessary to “accelerate our significant investments in AI and automation” and enhance the company’s technology capabilities.
Apple fixes Compressor code execution flaw
Apple released the Compressor 4.11.1 update for macOS Sequoia 15.6 and later to address a vulnerability (CVE-2025-43515) that could allow an unauthenticated user on the same network to execute arbitrary code. The security issue was mitigated by modifying the software to now refuse external connections by default.
Google reports 1000x reduction in Android memory bugs with Rust
Google’s Android team reported that using the Rust programming language has led to a 1000x reduction in the density of memory safety vulnerabilities compared to C and C++ code. The shift to Rust has made the secure development path faster, with Rust changes requiring 25% less time in code review and having a 4x lower rollback rate than C++.
EchoGram attack undermines AI guardrails
HiddenLayer researchers have uncovered EchoGram, a new attack technique that undermines common AI defense mechanisms like text classification and ‘LLM-as-a-judge’ guardrails. The exploit uses specific token sequences to manipulate the defensive model’s verdict, allowing malicious prompts to be approved or causing false alarms. This systemic vulnerability affects defenses used in major models like GPT-4, Gemini, and Claude.
Related: In Other News: Controversial Ransomware Report, Gootloader Returns, More AN0M Arrests
