With the new year just around the corner, you’re probably wondering what the latest cybersecurity threats will have in store for small businesses like yours. One such threat is the rise of agentic AI, which capitalizes on the weakest link in any business’ cybersecurity infrastructure: its human elements. If you already have a hard time figuring out if the person on the other end of the phone line is human, just wait… It’s only going to get worse.

However, that doesn’t mean you’re powerless.

Understanding Agentic AI

Agentic AI has a simple definition: the use of autonomous systems that are intelligent enough to perform multiple steps without human intervention.

It’s thought that agentic AI will be at the forefront of cybercriminal activity as they exploit it against business owners. Cybercriminals will be able to create hyper-realistic, real-time deception that they can deploy at massive scale. It’s thought these attacks will become serious enough to bring about a crisis of identity and authenticity in the business world at large. Seriously. It’s tough to conduct business when you can’t trust the person on the other end of the line.

Some strategies agentic AI might deploy include:

• AI-enabled deepfake social engineering - With real-time, flawless voice cloning (or vishing) and realistic text emulation, cybercriminals can take on the persona of CEOs or IT staff. It’s thought these deepfakes will be convincing enough to bypass multi-factor authentication, require wire transfers, or even trick your employees into running malicious code.
• Machine identities - There are several non-human identities on your infrastructure, by which we mean the automated scripts, cloud functionality, and application programming interfaces that keep your company operational. If a human-forged identity can infiltrate these automated systems, your cybersecurity defenses will come tumbling down simply by virtue of being recognized as a trusted entity.
• Prompt injection - If your business uses an LLM, it could become subject to prompt injection attacks. Attackers can manipulate the AI model to sneak through its security measures and execute malicious code. The LLM can also be corrupted into providing the attacker with sensitive data or run an application without the user’s permission.

While this all sounds seriously bad, don’t despair. We have some recommended strategies for you to try:

Here’s What To Do About Agentic AI

Your business has its work cut out for it, especially if you want to keep your company and your staff knowledgeable about these developing threats. For starters, you can ditch the reactive approach and adopt an identity-first security model. You’ll want to implement the following measures:

• Reconsider the SMS and one-time passcode model - Instead, choose a trusted 2FA app and ensure MFA is deployed everywhere possible on your infrastructure.
• Build zero-trust policies for AI agents - Make sure proper identity and access management controls are in place. This will help to track and audit any activities carried out by autonomous processes.
• Develop a crisis of authenticity response plan - In the event some situation does occur, you’ll need a plan in place. For example, your team should know how to respond when a bigwig suggests a large financial transfer or attempts to make other major decisions that have a high impact on the business. Any scenarios that involve voice and video should also be highly scrutinized and never trusted by default.

The cybersecurity landscape is changing, and you need the capacity to change along with it. Learn how to do so by working with First Column IT. Our professional technicians will walk you through the steps required along the way. Learn more by calling us at (571) 470-5594.

‍