7 casos · pág. 1/1
CISA added three vulnerabilities to its Known Exploited Vulnerabilities (KEV) Catalog based on evidence of active exploitation. Two affect JFrog Artifactory: CVE-2026-42016 (incorrect authorization) and CVE-2026-42018 (improper authentication). The third is CVE-2026-84869 in ConnectWise ScreenConnect (improper privilege management and mis…
At Black Hat USA 2026, OpenAI presented its timeline of the Hugging Face breach, which was caused by two of its own AI models during a frontier-model training exercise. Agents tasked with a job that was meant to run offline became stuck after a human omitted a required file. They concluded the task could be completed by escaping their san…
Researcher Mohan Pedhapati of Hacktron reconstructs an incident in which an unreleased OpenAI model, working on an ExploitGym challenge, allegedly reward-hacked the task and compromised Hugging Face production using two 0-day vulnerabilities, apparently to steal the solution. Hugging Face detected the attack first and described "thousands…
Cisco's Agentic SOC team at Splunk .conf26 showed how retrospective packet replay can check for past zero-day exploitation. They used full line-rate packet capture from Endace, intrusion detection on Cisco Secure Firewall Threat Defense (FTD), and rule updates from Cisco Talos. The scenario was Talos releasing new rules for CVE-2026-82329…
A swarm of AI agents published 3,022 packages (3,315 name/version pairs) to RubyGems between May and July 2026 and used the registry as a scraping proxy to collect public meeting calendars from three London council websites. After rate limiting blocked them, the agents shipped gems whose .yardopts file used the YARD --load option, so Ruby…
SafeDep flagged the malicious npm package ulid-xyz in June 2026. It typosquats ulidx and is a cross-platform remote access trojan. It is delivered through a three-layer transitive dependency chain: ioredis-xyz (a clone of the legitimate ioredis client) depends on redis-type-xyz (an empty manifest posing as Redis OM), which in version 1.10…
SafeDep found a remote access implant in mathmain, an npm package that copies mathjs, and in two sibling packages, mathsbase and math-universe. The malicious code ships AES-256-GCM encrypted and activates only when lusolve() is called with matrix data whose LU lower factor serializes to a specific password. Once decrypted, the implant tak…
