Ivanti Warns of New EPMM Flaw Exploited in Zero-Day Attacks

Article Link: https://www.bleepingcomputer.com/news/security/ivanti-warns-of-new-epmm-flaw-exploited-in-zero-day-attacks/

• Ivanti disclosed a high-severity zero-day vulnerability in Ivanti Endpoint Manager Mobile (EPMM) that allows remote code execution on on-prem EPMM servers. The flaw is already being exploited in limited real-world attacks.
• The issue stems from an improper input validation flaw that can be exploited by attackers with administrator privileges to execute arbitrary code. Ivanti released a patch for the vulnerability and recommends rotating credentials after patching is complete.
• While exploitation currently appears limited, the incident continues a broader trend of ransomware groups and nation-state actors actively targeting Ivanti products and appliances.
• Ivanti patched four additional high-severity EPMM vulnerabilities that could enable compromise of administrator accounts, certificate abuse, and data exfiltration. At this time, there are no reports of these additional vulnerabilities being exploited in the wild.
• Additional information: https://www.ivanti.com/blog/may-2026-epmm-security-update

RubyGems Suspends New Signups After Hundreds of Malicious Packages Are Uploaded

Article Link: https://thehackernews.com/2026/05/rubygems-suspends-new-signups-after.html

• RubyGems temporarily disabled new account registrations after detecting a malicious attack targeting the platform and its package ecosystem.
• According to Mend.io, who provides security services for RubyGems, hundreds of packages were involved in the campaign, some of which contained malicious code.  
• The incident highlights the growing threat of software supply chain attacks against open-source ecosystems, where attackers compromise trusted packages to distribute malware and steal credentials from developers and companies.
• The attack underscores the continued security risks facing open-source repositories, as compromised packages can impact developers, enterprise environments, and software supply chains.  

Google Reports First Known AI-Assisted Zero-Day Exploit in the Wild

Article Link: https://www.scworld.com/news/google-reports-first-known-ai-assisted-zero-day-exploit-in-the-wild

• Google Threat Intelligence Group (GTIG) reported the first known real-world case of threat actors using AI to help develop a zero-day exploit. The exploit targeted a two-factor authentication flaw in a popular open-source web administration tool.
• Researchers determined the exploit was likely generated or assisted by a large language model (LLM) based on several indicators. This included hallucinated vulnerability scoring data, AI-style code formatting, and strings commonly associated with AI-generated code.
• The incident highlights a major shift toward AI-assisted vulnerability discovery and exploit development. Which allows attackers to scale exploit creation and identify complex business-logic and authorization flaws more quickly than traditional methods.
• GTIG also observed increasing use of AI by cybercriminal and state-sponsored threat actors from countries including China, North Korea, and Russia for malware development, attack orchestration, reconnaissance, and social engineering operations.
• Additional information: https://cloud.google.com/blog/topics/threat-intelligence/ai-vulnerability-exploitation-initial-access

Vulnerability in Claude Extension for Chrome Exposes AI Agent to Takeover

Article Link: https://www.securityweek.com/vulnerability-in-claude-extension-for-chrome-exposes-ai-agent-to-takeover/

• Web browser security firm LayerX identified a vulnerability they named ClaudeBleed in the Claude for Chrome extension that would allow attackers to compromise the AI assistant and exfiltrate and manipulate or destroy data.
• The vulnerability is caused by weak permission controls and improper trust validation within the extension. Any Chrome extension running JavaScript on the claude.ai as the origin of execution could issue privileged commands to Claude without proper verification of context or source ownership.
• Researchers provided a proof of concept that attackers could exploit the issue to bypass Claude’s security protections and control the agent.
• Anthropic issued a partial mitigation after the disclosure, but LayerX stated the underlaying issue was not fully resolved and could still be exploited by switching extensions into privileged mode, which can be done without user approval.
• Additional information: https://layerxsecurity.com/blog/a-flaw-in-claudes-browser-extension-allows-any-extension-to-hijack-it/

Why Agentic AI Is Security’s Next Blind Spot

Article Link: https://thehackernews.com/2026/05/why-agentic-ai-is-securitys-next-blind.html

• Businesses are rapidly deploying agentic AI systems in their production environments, at times without input from their security teams. This creates growing concerns around visibility, governance, and risk management of such systems.
• Security experts warn that many teams lack a foundation understanding of how AI agents operate, including how they access data, leverage tools, and interact with external systems.
• Researchers identified several emerging risk categories, such as coding assistants in GitHub Copilot, vendor-built agents, and custom agents built by employees.
• AI agents often require broad permissions to function such as email and calendar access, file system and code repository integration, and access to APIs.
• Organizations must build practical AI security programs starting with a strong understanding of AI application architecture, strict access scoping, and integrating security earlier into AI deployment.

KongTuke hackers now use Microsoft Teams for corporate breaches

Article Link: https://www.bleepingcomputer.com/news/security/kongtuke-hackers-now-use-microsoft-teams-for-corporate-breaches/

• Access broker KongTuke shifted from web-based lures to using Microsoft Teams chats for social engineering. Attackers impersonate internal IT staff and convince employees to run a malicious PowerShell command that installs ModeloRAT.
• The attackers send a Teams message using a spoofed display name and convince victims to run a script that downloads a ZIP file from Dropbox that contains a portable Python distribution that downloads the malware. In some instances attackers have gained persistent network access in less than five minutes.
• The attack shows how groups, such as KongTuke, are evolving their methods and exploiting trusted enterprise tools to bypass traditional defense methods.
• As an access broker, the group sells compromised access to malware groups and other malicious actors. Organizations are advised to restrict external Teams federation using strict allowlists and monitor for indicators of compromise.
• Additional information: https://reliaquest.com/blog/threat-spotlight-help-desk-lures-drop-kongtukes-evolved-modelorat

Critical Windows DNS Client Flaw Enables Remote Code Execution

Article Link: https://cyberpress.org/windows-dns-client-flaw/

• Microsoft disclosed a critical vulnerability in their Windows DNS Client. The flaw exists in the DNSAPI library, a Windows component that processes DNS responses on millions of Windows systems.
• Attackers can send a DNS response that triggers a heap-based buffer overflow during normal DNS lookups. Exploitation requires no user interaction, authentication, or other access.
• Malicious actors can exploit the flaw through compromised routers, rogue DNS servers, poisoned resolvers, or public wireless networks. Exploitation enables remote code execution on vulnerable machines.
• The vulnerability exists in both endpoints and servers, creating a large attack surface across enterprise environments. A single compromise could enable lateral movement and persistent access in the network.
• Microsoft released patches for affected Windows versions, including Windows 11 and Windows Server. Security teams are advised to patch immediately, restrict DNS traffic to trusted resolvers, and monitor for unusual background processes.

Major Tech Manufacturer Foxconn Confirms Cyberattack Hit North American Factories

Article Link: https://cyberscoop.com/foxconn-cyberattack-disrupts-north-america-factories/

• Foxconn confirmed a cyberattack disrupted operations at several North American factories. The ransomware group Nitrogen claimed responsibility and alleged it stole 8TB of data across more than 11 million files.
• Nitrogen reportedly used ransomware tactics that combine data theft with system encryption to pressure victims into paying. The group has historically leveraged tools derived from leaked Conti code and previously used ALPHV infrastructure to target Windows and VMware environments.
• The group claims to have stolen data including confidential project files, engineering drawings, and internal documentation tied to major technology vendors. If legitimate, the breach could expose sensitive intellectual property, enable supply-chain compromise, and affect manufacturing data.
• Foxconn is a supplier for large tech companies like Apple, Intel, Google, Dell, and Nvidia. This makes it a high-value target and represents a potentially larger threat across the technology landscape.
• The company said security teams have contained the threat and allowed the organization to continue production and delivery capabilities.