Researchers have discovered a new data injection attack targeting AI agents, allowing attackers to force misclicks or execute arbitrary commands. The attack exploits how AI agents process untrusted data from external sources. This matters because it exposes a critical vulnerability in autonomous AI systems that could be exploited to compromise security.

Zoom has released a security update for a critical vulnerability in its Windows client that could enable attackers to take over user accounts. The flaw, if exploited, could allow unauthorized access to sensitive data and control of the affected system. Users are urged to update their software immediately to prevent potential compromise.

A new malware framework called OkoBot is injecting phishing prompts into legitimate Ledger and Trezor cryptocurrency wallet applications. The malware intercepts user interactions to steal seed phrases, potentially leading to the complete loss of digital assets. This attack matters because it undermines trust in hardware wallet security and poses a serious risk to users of these widely-used devices.

Security researchers highlight that SASE solutions are failing to address AI-generated threats, as traditional packet inspection cannot detect sophisticated attacks. This gap leaves organizations vulnerable to AI-driven malware and phishing. As cybercriminals increasingly leverage AI, security frameworks must evolve beyond packet inspection to incorporate AI-aware defenses.

Attackers compromised several AsyncAPI npm packages to deliver multi-stage botnet malware. Developers who installed these packages unknowingly integrated malicious code into their systems. This incident highlights ongoing supply chain risks in open-source ecosystems.

Microsoft released a record number of security updates, patching between 570 and 622 flaws—depending on the count—including two zero-day vulnerabilities that are currently being actively exploited. The unprecedented patch volume underscores the growing threat landscape and the urgent need for users to apply updates immediately.

SAP has released a security patch for a critical vulnerability in its NetWeaver ABAP platform, rated 9.9 on the CVSS scale. The flaw could allow attackers to expose or modify sensitive data without authentication. Given SAP's widespread use in enterprise systems, this patch is urgent for affected organizations.

Security vulnerabilities in RabbitMQ could allow attackers to leak OAuth authentication secrets and expose queue metadata across different tenants. The flaws affect users of the open-source message broker, particularly in multi-tenant deployments. This matters because it could lead to unauthorized access to sensitive data and breach tenant isolation.

Researchers discovered that 11 outdated Microsoft-signed Linux UEFI shims can be exploited to bypass Secure Boot protections. Attackers could use these to load unsigned bootloaders or malware, compromising system security. This matters because it affects millions of devices relying on Secure Boot for trusted startup.