
Claude Mythos 5 Tried to Backdoor a Project in UK AI Security Test
A Claude Mythos 5 agent spent 34 hours trying to merge malware into an open-source project during a UK AI

A Claude Mythos 5 agent spent 34 hours trying to merge malware into an open-source project during a UK AI

Google removed three ADK AI workflows after Pillar Security showed a GitHub issue could prompt-inject a triage agent into launching

cPanel patched CVE-2026-58048, a CVSS 9.4 privilege-escalation flaw letting an authenticated hosting customer execute SQL in the database root context.

Unit 42 found XCSSET v40 targeting macOS developers via compromised Xcode projects, adding a Chrome hijacker and Telegram trojanizer to

A Google Firebase misconfiguration in the tl;dv AI meeting tool lets users query others’ meeting data and potentially join calls,

Socket found 18 malicious npm packages impersonating Alibaba developer tools that deliver a cross-platform RAT with remote control and data-theft

Threat actors poisoned Xanadu’s mrmustard 0.7.4 on PyPI with an info-stealer that exfiltrates SSH keys and AWS credentials from research

FaceHugger flaws in Hugging Face Diffusers bypass trust_remote_code and let malicious model repositories execute arbitrary code when models are loaded.

Attackers tampered with Adform’s trackpoint script, rewriting crypto wallet addresses across customer pages to divert payments to attacker-controlled wallets.

Wiz researchers showed an Azure Cosmos DB Gremlin sandbox escape could expose a platform-wide signing key that unlocks any tenant
Subscribe to the Daily Security Review Email Briefing to stay informed on the latest threats, trends, and technology, along with insightful columns from industry experts.