AI & software security
-
Meta Muse AI agent appears as ordinary browser traffic, Cequence research finds
Cequence observed traffic matching Meta’s Muse at more than half its sampled customers, including authenticated account activity that conventional browser checks could not distinguish from human use.
-
Microsoft Teams prepares integrations to detect deepfakes and meeting impersonation
Microsoft is developing support for external deepfake-detection providers in Teams, reflecting growing concern about identity verification during enterprise video meetings.
-
GitHub adds contextual AI detection for leaked passwords and developer secrets
GitHub has introduced a specialised model intended to identify credentials that conventional secret-scanning patterns can miss, with different features at different rollout stages.
-
GitHub expands local sandbox controls for Copilot coding agents
GitHub has announced broader availability of local sandboxing to restrict the files, credentials and networks that coding agents can reach while executing tools.
-
Cisco patches critical flaws in on-premises licence management software
Cisco has issued critical fixes for License On-Prem systems, including a security hardening release with a maximum CVSS score and no available workarounds.
-
Pwn2Own Ireland researchers demonstrate weaknesses in connected and AI systems
Researchers competing in Cork demonstrated vulnerabilities across connected devices and AI infrastructure, with coordinated disclosure intended to give vendors time to release fixes.
-
Tensorlake compromise exposes AI development infrastructure to credential theft
A compromised release of the Tensorlake software development kit installed credential-stealing malware through an npm lifecycle script, exposing systems used to develop and run AI agents before their isolated e
-
Traefik fixes eight vulnerabilities across proxy software
Traefik has released fixes for eight security vulnerabilities capable of affecting data confidentiality, integrity, security controls and service availability.
-
WordPress 7.1.3 fixes seven security flaws
WordPress 7.1.3 fixes seven security weaknesses including stored cross-site scripting, SQL injection and unauthenticated disclosure of comments attached to private or unpublished posts.
-
Cisco fixes six NX-OS vulnerability classes
Cisco’s October NX-OS hardening release fixes six CVEs across its switching software, including access control and memory corruption classes carrying maximum CVSS scores of 9.8.