Cloud application hosting platform Northflank is warning customers to rotate all credentials and secrets stored on its platform after a security breach led to unauthorized use of customers' AI API keys. [...]

A security researcher has released a proof-of-concept exploit for a recently patched Android vulnerability, dubbed "SMS of Death" and tracked as CVE-2026-58882, that lets attackers remotely disrupt cellular service or crash phones by sending a specially crafted text message. [...]

Reddit announced in a post by co-founder Steve Huffman that the site will gradually become a "logged-in service," offering limited access to visitors without accounts. [...]

AI-generated vulnerability reports are overwhelming some bug bounty programs, making it harder for aspiring security researchers to know where to start. TryHackMe explains how beginners can build practical cybersecurity skills, use AI responsibly, and produce reproducible findings that stand up to scrutiny. [...]

Blockchain analytics firm Chainalysis disclosed a data breach after attackers used "offensive agentic techniques" to access internal systems and tools, exposing customer names, email addresses, user IDs, and organizational roles. [...]

Cybersecurity professionals are calling the rapid emergence of AI-detected vulnerabilities “Black October,” with some flaws appearing faster than organizations can patch them. Meanwhile, AI coding tools are helping attackers target banks and other institutions, raising fears that a major incident could arrive before the industry coordinates stronger safeguards.

Source

Microsoft Execution Containers (MXC) sandbox AI agents, limiting which files and network resources they can access; process-container support spans Windows 11, macOS, and Linux. Microsoft is also expanding Project Perception with agent-specific identities, task-scoped permissions, and controls that can tighten access or stop an agent as its risk changes. The goal is to govern agents as distinct actors—not simply trust the user accounts they operate under.

Source

Microsoft says organizations should treat attackers probing OneDrive URLs for usernames as a low-risk threat, not a sign that accounts have been breached. The probes can reveal user principal names (UPNs), which attackers might later use for phishing, credential stuffing, or social engineering. Microsoft recommends strong identity protections to limit that downstream risk.

Source

Only 13% of organizations believe they have adequate AI-agent governance, according to Gartner data cited by IBM. Microsoft’s guidance gives identity teams a practical starting point: strengthen authentication, remove standing privileges, and apply consistent access policies so agent identities receive least-privilege protections.

Source

Microsoft reports two intrusions that turned Teams help-desk impersonation and user-approved Quick Assist sessions into attacks on Active Directory Certificate Services (AD CS), which issues certificates used for authentication. One attacker abused an ESC1 template and continued using a privileged account excluded from automated containment to steal data. In a separate ESC8 attack, isolating the affected workstation stopped further observed activity.

Source

Security researchers at Anthropic have developed an exploit for a critical Linux kernel vulnerability that allows unauthenticated attackers to remotely execute commands as root on servers running Multipath TCP (MPTCP) listeners. [...]

Synology Ransomware Alert – Lock3D, C3RB3R, PHP2SHELL and Synology’s BIG User Problem Synology NAS owners are currently dealing with 2 separate security stories. China-based cybersecurity company Qihoo 360 says its anti-ransomware service received multiple requests for help since early October from people whose Synology devices had been hit by C3RB3R, a variant of the Cerber […]

Hardware wallet device maker Trezor has received credible reports about a suspected hardware crypto-theft implant in a Trezor Safe 5 sold by its reseller CryptoBilis in Southeast Asia. [...]

Claude Code’s VS Code extension brings Anthropic’s coding assistant to compatible editors. A tampered version, 2.1.313, was available through Cursor’s extension marketplace for about 11 hours on October 29, 2026, and attempted to install persistent SSH access on affected computers. Cursor says the attacker’s server is now offline, but affected users still need to check and clean their systems.

Source

A malicious version of Anthropic's Claude Code VS Code extension was distributed through Cursor's extension marketplace for approximately 11 hours on October 29, exposing up to 5,000 developers to malware designed to install persistent SSH backdoors. [...]

Flaws in SS7 and Diameter signaling protocols can allow attackers to track phone locations, intercept calls and SMS verification codes, and impersonate subscribers without compromising their devices. Cape explains how these attacks exploit trust between networks and how its security-focused mobile service helps defend against them. [...]

Attackers have started exploiting a maximum-severity GitLab vulnerability patched earlier this month, according to the US Cybersecurity and Infrastructure Security Agency (CISA). [...]

Mistral Large 4 is a multimodal AI model for coding, cybersecurity, and other work that organizations can run through hosted services or deploy on their own infrastructure. Mistral has now released its weights after three more weeks of training improved coding results and reasoning efficiency. The model leads open-weight rivals on several coding benchmarks, while a specialized security variant is available to vetted enterprise partners.

Source

AI attackers clicked on planted decoys far more often than expert human hackers in Horizon3 research, even when the models recognized a possible trap. That finding challenges claims that frontier models can already take over attacks end to end: they can multiply attempts, but deception can expose them. For defenders, the more immediate risk may be AI speeding up familiar attacks, including exploiting vulnerabilities before organizations patch.

Source