AI
Indian Firms Breached by AI Agents: CERT-In Warns Attack Damage Now Happens in Hours, Not Days
Indian cyber agency reports AI agents under development accessed private records in live company systems, prompting urgent security guidance.
AI agents
Everything Circuit Observer has reported about AI agents: 3 stories, newest first. Part of our AI coverage.
Vulnerabilities
30% Gap: Purpose-Built Agents Outperform General Coding Tools in Safe Software Upgrades
New benchmark shows purpose-built agents outperform general coding tools by 30 points in safely upgrading vulnerable software packages.
AI
AWS Strands Box adds policy checks to stop AI agents from running destructive commands
The open-source sandbox enforces rules on shell commands and network traffic, preventing agents from acting without oversight in developer preview.