A new study shows that verse-based prompts can slash the effectiveness of AI safety constraints. We’re breaking down an experiment involving 25 language models and its key takeaways.
A new ransomware called Osiris was used in an attack on a major food service franchisee operator in Southeast Asia in November 2025. The ransomware shares similarities with previous Inc ransomware attacks, including the use of Wasabi buckets for data exfiltration and a specific version of Mimikatz. Osiris has typical ransomware functions, uses a hybrid encryption scheme, and drops a ransom note. The attack chain involved data exfiltration using Rclone, deployment of dual-use tools, and the use of a malicious driver called Abyssworker or Poortry. The attackers employed bring-your-own-vulnerable-driver (BYOVD) techniques to disable security software. While the impact of Osiris on the ransomware landscape remains uncertain, it appears to be wielded by experienced attackers with potential links to Inc ransomware or its affiliates.
This intelligence report emphasizes the importance of understanding one's own network environment and not ignoring reconnaissance events in cybersecurity. It highlights the increasing sophistication of bad actors in reconnaissance, both in network scanning and social engineering, aided by AI tools. The report warns against dismissing reconnaissance alerts in favor of focusing solely on attack signals, stressing that initial access brokers excel at understanding target environments. Recent vulnerability discoveries in various software applications are mentioned, along with key security headlines including phishing campaigns, ransomware attacks, and nation-state hacking activities. The report also provides information on prevalent malware files and upcoming security events.
KONNI, a North Korean-linked threat actor active since at least 2014, has evolved its tactics by adopting AI to generate sophisticated PowerShell backdoors. This ongoing phishing campaign targets organizations primarily in South Korea but could potentially impact European entities involved in diplomacy, international relations, NGOs, academia, and government sectors. The AI-generated backdoors increase the complexity and variability of malware, making detection and mitigation more challenging. Although no known exploits in the wild have been reported yet, the medium severity rating reflects the potential risk posed by this advanced malware. European organizations with strategic ties to South Korea or those in sensitive sectors should be vigilant. Mitigation requires enhanced email filtering, behavioral monitoring of PowerShell usage, and threat intelligence sharing. Countries with significant diplomatic, academic, or NGO presence related to Korean affairs, such as Germany, France, and the UK, are more likely to be targeted. The threat is medium severity due to the moderate impact on confidentiality and integrity, the complexity of exploitation, and the lack of requirement for user interaction beyond phishing. Defenders should focus on AI-driven malware detection capabilities and proactive phishing defense measures.
This comprehensive analysis covers cyber threats and security issues in the financial industry, both in Korea and globally. It examines malware and phishing cases, lists top malware strains, and provides statistics on leaked Korean accounts. Key issues on the deep and dark web are highlighted, including a major database leak from Indonesia's largest bank, exposing sensitive financial data of approximately 3 million customers. A ransomware attack on a leading African financial services company by INC Ransom group is also detailed, with 100GB of data reportedly stolen. The report emphasizes the potential for widespread damage and chain attacks, urging proactive measures among financial institutions and related companies.
A comprehensive analysis of a Potentially Unwanted Application (PUA) identified as HEURRemoteAdmin.GoToResolve.gen reveals its association with the GoTo Resolve Unattended Access application. While digitally signed by GoTo Technologies USA, LLC, the sample exhibits behaviors typical of PUAs, including silent installation, background thread execution, and persistent presence on the system. The application's use of the Restart Manager library, often seen in ransomware and wiper malware, raises concerns. Although no direct malicious payload was observed, the remote access capabilities present a significant security risk, potentially allowing unauthorized system control or deployment of secondary malware. The sample's detection by UltraAV further supports its classification as a security threat, warranting removal unless explicitly authorized and managed within organizational security policies.
API cybersecurity will be a ping pong ball, battered between the rackets of AI-assisted attackers and AI-assisted defenders. The post Cyber Insights 2026: API Security – Harder to Secure, Impossible to Ignore appeared first on SecurityWeek .
LockBit, originating as ABCD ransomware in 2019, has evolved to version 5.0 in September 2025. After a period of inactivity, it resumed operations in December 2025 with a reduced affiliate sign-up fee. LockBit 5.0, nicknamed ChoungDong, consists of a Loader and Ransomware component. The Loader decrypts and executes the payload in memory, while the Ransomware uses ChaCha20 and Curve25519 for encryption. This update significantly enhances evasion techniques and attack efficiency, introducing features like Mutex, Execution Delay, and Wiper. The group's history includes affiliation with the Maze cartel, independent operations, and continuous upgrades. Mitigation strategies involve monitoring process behavior, applying security patches, and preparing for swift responses using provided IoCs and MITRE ATT&CK techniques.
Visual Studio Code is a popular open-source code editor[1]. But it's much more than a simple editor, it's a complete development platform that supports many languages and it is available on multiple platforms. Used by developers worldwide, it's a juicy target for threat actors because it can be extended with extensions.
Under the new rules, measures for 5G cybersecurity would become mandatory. The post EU Plans Phase Out of High Risk Telecom Suppliers, in Proposals Seen as Targeting China appeared first on SecurityWeek .
Providing cyberespionage and remote code execution capabilities, the malware is executed via DLL sideloading. The post APT-Grade PDFSider Malware Used by Ransomware Groups appeared first on SecurityWeek .
A simple payload allowed attackers to create a new event leaking summaries of the victim’s private meetings. The post Weaponized Invite Enabled Calendar Data Theft via Google Gemini appeared first on SecurityWeek .
VoidLink is an advanced AI-generated malware framework exhibiting high maturity, modularity, and sophisticated techniques such as eBPF and LKM rootkits. It targets cloud and container environments with dedicated modules for enumeration and post-exploitation, indicating a focus on modern infrastructure. Although no known exploits in the wild have been reported yet, its architecture suggests a significant threat potential. European organizations relying on cloud-native and containerized deployments are at heightened risk due to the malware's capabilities. Mitigation requires advanced detection mechanisms tailored to eBPF and kernel-level rootkits, alongside strict cloud environment monitoring and segmentation. Countries with strong cloud adoption and critical infrastructure sectors, such as Germany, the UK, France, and the Netherlands, are likely to be most affected. Given its high impact on confidentiality, integrity, and availability, combined with ease of exploitation via AI-generated code, the suggested severity is high. Defenders should prioritize proactive threat hunting and behavioral analytics to detect this evolving threat.
US officials told The New York Times that cyberattacks were used to turn off the lights in Caracas and disrupt air defense radars. The post New Reports Reinforce Cyberattack’s Role in Maduro Capture Blackout appeared first on SecurityWeek .
A data breach was disclosed by Spanish energy company Endesa following unauthorized access to a commercial platform managing customer information. Attackers reportedly exfiltrated over 1 terabyte of data, including sensitive financial details such as IBANs. This incident highlights risks to critical infrastructure sectors in Europe, particularly energy providers. The breach's medium severity reflects significant confidentiality impact but no known active exploits or public patches yet. European organizations, especially in energy and utilities, face potential risks from similar unauthorized access attacks. Mitigation should focus on strengthening access controls, continuous monitoring, and rapid incident response tailored to commercial platforms handling sensitive customer data. Spain and neighboring European countries with interconnected energy markets are most likely affected due to shared infrastructure and market penetration. Given the data volume and sensitivity, the threat severity is assessed as high, considering the potential for financial fraud and reputational damage. Defenders should prioritize securing customer data platforms and enhancing threat intelligence sharing within the sector.
PDFSIDER is a newly identified malware variant that utilizes DLL side-loading to deploy a covert backdoor with encrypted command-and-control capabilities. It exploits vulnerabilities in legitimate software like PDF24 Creator to bypass endpoint detection mechanisms. The malware operates primarily in memory, minimizing disk artifacts, and employs advanced anti-VM technology to evade sandboxes and analysis labs. PDFSIDER features a robust cryptographic implementation using the Botan library for secure communications. It gathers system information and provides attackers with an interactive, hidden command shell for remote execution. The malware's characteristics align with APT tradecraft, suggesting its use in cyber-espionage operations. Distribution occurs through spear-phishing emails containing ZIP archives with legitimate-looking executables.
