← Back
VulnerabilityUnit 42·5 hours ago

Perturbation Probing: A New Diagnostic for the Fragility of LLM Safety

New research reveals that AI safety refusal lives in a thin neural layer, highlighting the critical need for external, multi-layered security. The post Perturbation Probing: A New Diagnostic for the Fragility of LLM Safety appeared first on Unit 42.

Read full article at Unit 42

Related Articles

VulnerabilityTenable·13 hours ago

Why a cryptographic inventory is key for addressing the quantum computing threat

Adversaries are already harvesting encrypted data today through "harvest now, decrypt later" tactics, planning to decrypt it once quantum computers mature, making the quantum threat an immediate operational concern rather than a distant future problem. Organizations must establish comprehensive cryptographic inventories and execute a phased migration strategy to quantum-resistant algorithms, with regulatory mandates including Executive Order 14412 setting federal deadlines of 2030-2031 for transitioning to post-quantum cryptography. Without complete visibility into cryptographic assets across their environment, organizations cannot effectively prioritize and remediate quantum vulnerabilities before attackers can exploit them with future quantum capabilities.

VulnerabilitySecurity Affairs·13 hours ago

U.S. CISA adds ownCloud, Linux Kernel, and JFrog Artifactory flaws to its Known Exploited Vulnerabilities catalog

CISA has added three actively exploited vulnerabilities to its Known Exploited Vulnerabilities catalog, including a critical authentication flaw in ownCloud Server's WebDAV functionality with a CVSS score of 9.8, alongside flaws in the Linux Kernel and JFrog Artifactory. These additions to the KEV catalog indicate that threat actors are actively exploiting these vulnerabilities in the wild, making them priority targets for patching across affected organizations.

VulnerabilitySecurityWeek·14 hours ago

OpenAI Agents Exploited Linux Kernel Flaw on Company’s Own Systems

CISA has added the exploited flaw, CVE-2026-53362, to its KEV catalog, alongside a JFrog vulnerability exploited by OpenAI agents. The post OpenAI Agents Exploited Linux Kernel Flaw on Company’s Own Systems appeared first on SecurityWeek.

VulnerabilityThe Hacker News·15 hours ago

Two Unitree G1 EDU Humanoid Robot Flaws Enable Root RCE, One Starts Over Bluetooth

Security researcher Olivier Laflamme discovered two independent root RCE vulnerabilities in the Unitree G1 EDU humanoid robot, tracked as CVE-2026-76639 and CVE-2026-76640. One attack path exploits chat_go and bashrunner over the network, while the other leverages Bluetooth Low Energy to achieve root access on the robot's Locomotion PC.