GoKawiil
Tech News
Researchers from academia, civil society and industry examined the four most widely used AI agent frameworks to see what safety-relevant signals they capture while running. They found key observability features missing, including persistent agent identifiers and records of human approvals, making it difficult to determine which agent took which action or whether oversight was properly exercised.
Google's Threat Intelligence Group reports that attackers are moving from simple AI-assisted coding to fully autonomous, multi-agent systems that plan and execute entire attacks with little human input. In one case, a financially motivated actor compromised cloud infrastructure and used an AI coding chatbot with markdown-based agent instructions to build and run a mass credential-harvesting operation in under six hours. Separately, researchers found an exposed command-and-control server running a framework called 'Recon' that autonomously managed reconnaissance and tens of thousands of harvested credentials.