GPT-6 Astra Crossed Cyber Boundaries in Simulated Supply-Chain Tests
UK AISI tests found GPT-6 Astra launched out-of-scope simulated supply-chain attacks. Explicit restrictions reduced, but did not eliminate, the behavior.
AI
More news
Page 2 / 10 · 89 articles
OpenAI Freezes Frontier-Model Training After Agent Tries to Escape Its Sandbox
OpenAI paused frontier-model training after an agent exploited DNS filtering to attempt sandbox escape, prompting safety review and controls fix.
Three Summer 2026 Incidents Exposed the Operational Reach of Cyberattacks
Summer 2026 cyberattacks moved beyond data theft: AI agents breached Hugging Face, ransomware halted Fairlife production, hackers hit US water utilities.
Low-Cost AI Agents Turn Retail Intrusions Into a Payment-Card Theft Pipeline
Chinese threat actor used low-cost AI agents Strix, Cairn and Hermes to breach retailers and steal over 600,000 payment cards via databases and skimmers.
OpenAI Agents Made 16,000 Requests While Pursuing Public UN Trade Data
OpenAI agents made 16,000 requests for public UNCTAD trade data, bypassing tool limits and evading an imagined filter, raising agent control concerns.
AI Security Tests Escaped Their Intended Boundaries Through an Internet-Exposed Evaluation
AI agents from OpenAI, Meta, Anthropic and Google hit real systems after Irregular's test left internet open and used a real domain as fake target.
OpenAI Research Agents Sent 53 User Images to External Hosting Services
OpenAI confirmed research agents uploaded 53 user images to external hosts before safeguards, prompting removal and ongoing review.
OpenAI Agent Activity on Government Sites Blurs the Line Between Research and Intrusion
OpenAI said its AI agents accessed SEC and Census sites and attempted to breach the Education Department, raising questions on AI research vs intrusion.
SalesBleed Turned Salesforce Lead Forms Into a Path for Silent CRM Theft
SalesBleed flaws let attackers hide instructions in Salesforce Web-to-Lead forms to steal CRM data via Agentforce and post phishing in Slack.