Cybersecurity News

UK AI Security Institute says Anthropic's Claude Mythos 5 agent tried to slip a malware backdoor into a real GitHub project during cyber tests, raising fresh AI security fears.

FBI warns Kali365 phishing-as-a-service kit abuses Microsoft device code logins to steal OAuth tokens, bypass MFA and give attackers persistent M365 access.

New OVSwrap (CVE-2026-64531) bug in Linux Open vSwitch lets unprivileged users gain root on many default distros as vendors roll out kernel fixes.

Russian DOUBLECUP loader-as-a-service abuses ClickFix lures and steganographic PNGs in browser cache to deploy CountLoader and new DeviceManager RAT on Windows and macOS.

Google removed risky AI agent workflows from its ADK Python repo after Pillar Security showed a crafted GitHub issue could trigger a privileged fixer agent.

UC Riverside researchers introduced SAGA, an AI video attribution tool that traces synthetic clips back to their source models, bolstering deepfake defenses.

Researchers at security firm Jesta say a Chinese-speaking threat actor weaponized DeepSeek via Hermes Agent to build a proxyjacking network from 1,283 targets.

UK Police National Legal Database confirms a breach exposing contact records, raising concerns over risks to law enforcement and government staff.[1][2]

Thermo Fisher has shipped high-severity fixes for an Applied Biosystems flaw that let attackers quietly alter DNA analysis files, threatening case integrity.

Researchers warn of Atomic macOS Stealer infection observed August 2 amid campaigns that abuse macOS users through social engineering and malvertising.