Cyberthreats Outpacing SMBs: AI Widens the Attack Surface and Speeds Attacks

The pace of cyberthreat evolution has increased because AI both expands the attack surface and empowers threat actors to act faster. Small and medium businesses (SMBs) are adopting AI for customer service, development, finance, and other functions. That adoption creates new points of access to sensitive data and systems and often outstrips the governance needed to control permissions and behavior.

Public repositories of AI 'skills' and similar plugins are a growing risk vector. ESET analyzed roughly 900,000 skills and identified over 25,000 suspicious items and more than 3,000 tagged as malicious. These components can exfiltrate data, run malware, override instructions through prompt injection, or change the behavior of agents.

Prompt injection is especially dangerous because it turns content into an attack vector. Malicious instructions can be fed directly to models or hidden in content the model later reads, enabling attackers to manipulate agent behavior at scale.

AI also strengthens attackers' toolbox. It automates victim reconnaissance across social media and web sources, maps relationships for targeted phishing, and scales the pre-attack discovery process. AI assists in vulnerability research and exploit development, enabling attackers to find and weaponize flaws faster than defenders can patch systems. Social engineering becomes more convincing and localized as AI composes fluent, personalized messages.

For SMBs, the practical consequence is an exploitation window that has narrowed. Incidents that formerly unfolded at a human pace can now progress in hours or minutes. SMB security teams, already stretched thin and often responsible for a broad IT remit, lack time and budget to keep pace with this acceleration.

That creates a demand for a different operational model. Security solutions should automate routine detection and mitigation to reduce the workload on small teams. Automation needs to be combined with clear human oversight for decisions that require context and judgment. Finding partners that provide usable automation alongside expert support can help SMBs improve readiness and resilience without unsustainable costs.

Concrete focus areas for SMBs include establishing AI governance (policies, permissions, and lifecycle controls), vetting third-party AI components and repositories, hardening sources that feed agents, and improving incident response to operate at machine speed. The balance should be pragmatic: deploy automation where it saves time and reserve human review for high-risk or ambiguous cases.