AI Vulnerabilities
-
Anthropic may be preparing public rollout of restricted Claude Mythos model
Anthropic appears to be preparing a public rollout of its restricted Claude Mythos model after it briefly surfaced in Claude Code and Claude Security, following an April preview that said it could generate highly capable cyberattacks.
-
Mozilla says AI-assisted Mythos found 271 Firefox vulnerabilities with few false positives
Mozilla said its Mythos AI-assisted security research found 271 Firefox vulnerabilities, including 180 rated sec-high. The company faced skepticism over false positives and the lack of individual CVEs.
-
UK’s NCSC pilots Proactive Notifications to warn organisations of exposed devices
The UK’s National Cyber Security Centre has begun piloting Proactive Notifications, a Netcraft-delivered service that scans public internet data to warn organisations about exposed devices and recommend updates; it complements the NCSC’s Early Warning alerts but is not a replacement and has no announced end to the pilot phase.
-
Google issues December Android security updates, patches 107 flaws including two exploited in the wild
Google released December 2025 Android security patches that fix 107 vulnerabilities across multiple components, including two Framework flaws reported as exploited in the wild; users and manufacturers are urged to apply the 2025-12-01 or 2025-12-05 updates.
-
Canada warns of hacktivist breaches at water, energy and farm facilities
The Canadian Centre for Cyber Security warned that hacktivists have repeatedly breached internet-exposed industrial control systems at water, oil and agricultural sites, altering control settings and prompting guidance to remove direct internet exposure, use multifactor VPNs, and report incidents through the Cyber Centre.
-
Trend Micro: RondoDox botnet campaign expands to exploit more than 50 flaws across 30 vendors
Trend Micro said RondoDox campaigns have widened to exploit more than 50 vulnerabilities across over 30 vendors, using a loader-as-a-service model that bundles RondoDox with Mirai and Morte, and researchers linked the activity to large-scale botnet operations and coordinated RDP attacks.
-
Microsoft patches critical Entra ID flaw CVE-2025-55241 with 10.0 severity, enabling cross-tenant impersonation
Microsoft has patched a critical Entra ID vulnerability, CVE-2025-55241, with a maximum CVSS score of 10.0 that could allow cross-tenant impersonation of users, including Global Administrators. The fix, disclosed by researchers and implemented on July 17, 2025, requires no action by customers, though experts warn the flaw highlights broader cloud-security risks and the need to…
-
LNER confirms customer data accessed in third-party data breach
London North Eastern Railway says customer contact details and some journey information were accessed via a third-party supplier, with no impact on ticketing or services and no storage of bank data. The company urges caution on phishing and emphasizes secure passwords.
-
CISA Adds D-Link Wi-Fi Camera Vulnerabilities to Known Exploited Catalog Amid Active Threats
CISA has added three vulnerabilities affecting D-Link Wi-Fi cameras and video recorders to its KEV catalog amid warnings of active exploitation. Users are urged to take action as fixes for some issues have already been released.
-
Critical Vulnerabilities Discovered in NVIDIA’s Triton Inference Server
A set of critical vulnerabilities in NVIDIA’s Triton Inference Server has been discovered, posing significant risks to organizations using the platform for AI operations. Potential exploits could lead to remote control of servers and theft of sensitive data.
