Search results for LLMjacking
AI & Enterprise
Google report warns fully autonomous AI cyberattacks could emerge within 6 months
Google has confirmed for the first time a case in which attackers fully automated vulnerability scanning and exploitation using an LLM framework without human involvement, SiliconANGLE reported. Google Threat Intelligence Group disclosed a framework called exploit_agent found on an exposed test server and linked to multiple large language models. It compared web application code before and after patches, wrote attack code and ran it automatically. The system targeted French media, entertainment and education sites.
AI & Enterprise
AI, IT firms step up security as South Korea retools national security policy
Security threats around AI are intensifying, prompting more AI and IT companies to manage AI security directly through new gateways, patents and partnerships. Hacking incidents continued this week, including breaches at a South Korean financial cybersecurity body and certificate issuer DigiCert. The United States and France pointed to Iran over cyber attacks on supermarket refrigeration systems. South Korea is launching a cyber resilience task force, tightening IoT and cold-chain security, and raising crypto-asset operators’ standards.
AI & Enterprise
AI \'dark economy\' spreads as stolen computing power access is traded
Hacking aimed at stealing computing power is spreading as demand for AI grows, differing from attacks focused on personal data or ransomware. The Wall Street Journal reported that more than 10 AI development platforms were hit in the past month, with attackers targeting API credentials. Stolen keys have been used to access U.S. AI models, generating large fees and enabling resale through illicit routing services. Google analysts describe the tactic as LLMjacking.