Kaspersky will release a new tool later in August that checks AI agents, models and downloadable agent skills for malware, backdoors and bias before they reach company networks. The product aims to counter risks from shadow AI, where staff use chatbots, coding assistants and agent marketplaces without IT approval. Adrian Hia, managing director for Asia-Pacific at Kaspersky, announced the plan at the company's APAC Cyber Security Weekend. The tool will sit under existing endpoint and threat intelligence products. It will begin by scanning OpenClaw skills both at download and at the point of use. Kaspersky has already tested public agentic AI skills and found ongoing vulnerabilities. The company said it has identified more than 15,000 malware samples disguised as agentic AI software so far this year. Hia noted that AI agents create a new supply chain layer because they rely on third-party frameworks, APIs and plugins, so one compromised piece can affect many systems downstream.
Kaspersky will release a new tool later in August that checks AI agents, models and downloadable agent skills for malware, backdoors and bias before they reach company networks. The product aims to counter risks from shadow AI, where staff use chatbots, coding assistants and agent marketplaces without IT approval. Adrian Hia, managing director for Asia-Pacific at Kaspersky, announced the plan at the company's APAC Cyber Security Weekend. The tool will sit under existing endpoint and threat intelligence products. It will begin by scanning OpenClaw skills both at download and at the point of use. Kaspersky has already tested public agentic AI skills and found ongoing vulnerabilities. The company said it has identified more than 15,000 malware samples disguised as agentic AI software so far this year. Hia noted that AI agents create a new supply chain layer because they rely on third-party frameworks, APIs and plugins, so one compromised piece can affect many systems downstream.