The SMB cybersecurity squeeze: AI agents at work, old attacks in overdrive
ESET warns SMB AI agents face malicious skills and prompt injection while classic attacks still dominate breaches.
ESET describes how small and mid-size businesses are adopting multi-agent AI while security staffing stays thin. A survey of 4,400 SMB decision-makers found 40 percent had no AI policy, and policies were more common after an incident. Scanning almost 900,000 agent skills from March to May 2026 flagged more than 25,000 as suspicious and more than 3,000 as malicious, including cases leading to credential theft, data exfiltration, and remote code execution. The article also covers indirect prompt injection, citing EchoLeak in Microsoft 365 Copilot and a red-team contest that hijacked all 13 frontier models tested, while phishing and exploitation of known vulnerabilities remain the leading SMB breach causes.
- ESET surveyed 4,400 SMB leaders; 40 percent had no AI policy.
- Of nearly 900,000 agent skills, over 3,000 were outright malicious.
- Malicious skills enabled credential theft, data exfiltration, and remote code execution.
- Indirect prompt injection hijacked all 13 frontier models in a red-team contest.
- Phishing and known-vulnerability exploits remain the top SMB breach causes.
- Products
- Microsoft 365 Copilot
From the original article
8 min read at welivesecurity.com
As AI opens new paths to company data while making familiar attacks faster and cheaper, SMBs need protection designed around the time and expertise available to operate it
welivesecurity.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.