AI models are everywhere right now, popularized by LLMs, chatbots and conversational AI that sound more and more realistic every day. From Google Gemini to Apple’s less overt Apple Intelligencewe’re pushing the boundaries of what these complex algorithms can recognize and manufacture. That technology is hard at work in home security too and nothing shows it better than AI security cameras.

While your home camera isn’t going to strike up a conversation — not quite yet — it can identify and report a variety of objects with the latest built-in intelligence. Here’s how it all works.

Read more: I Thought I’d Hate AI in Home Security. I Couldn’t Have Been More Wrong

What does an AI home security camera do?

While AI security cameras won’t answer questions like ChatGPT, they use similar complex algorithms for another purpose: To categorize and identify objects. These learning algorithms are trained on many different images so that they can spot common things a security camera might see, like a dog, human or vehicle.

When paired with motion detection, this kind of object recognition can send smart alerts letting you know that specific things are happening — like when a package shows up at your front door or suddenly disappears. Or the cameras can automatically filter out passing vehicles so they aren’t triggered by every car driving by.

Deeper customization options are often hidden in camera settings, allowing you to arrange for specific types of object recognition or turn off the feature altogether. For example, you could set a camera to only send you alerts when it sees a human figure. But that’s only the beginning of what AI cameras are now capable of.

What are the advantages of having these AI models in home security?

In my experience testing dozens of different home security AI, the biggest everyday benefits come down to three important things:

• These kinds of smart alerts save time and limit random notifications
• AI allows you to focus on your home security on what’s important to you
• Home camera algorithms have the potential to solve specific problems, like keeping an eye on porch theft or only letting you know when strangers are approaching your home

Keep in mind, you don’t have to use AI features on your home security cams. Most have the ability to turn them off. But object detection is easy to use and has relatively few drawbacks.

Are there any disadvantages to using AI security cameras?

The biggest downside to this kind of AI is that new users will need to spend time in their camera app settings learning where object detection is, what they want to enable and how it all works. There’s a learning curve if you haven’t bought one of these cameras before.

Some versions of these algorithms also required subscription fees or may raise privacy concerns, issues I’ll touch on below.

Read more: This Major Blunder Shows Why You Can’t Trust ChatGPT with Home Security Questions

Can these algorithms recognize human faces?

Some can, and it’s becoming more common across brands like SimpliSafe, Google Nest and Arlo. Face recognition allows you to save profiles of family or people who regularly visit your home so that your camera can recognize specific people.

These face profiles let you receive even more targeted alerts, and in some cases can automate home security, like unlocking a smart lock when the right familiar face shows up. Tech like ADT’s Trusted Neighbor access (which uses Google Nest’s familiar face capabilities) can also give faces temporary access, or limit their access to only certain times of day — like when a dogwalker shows up. SimpliSafe’s version, meanwhile, can send alerts if there’s an unfamiliar face, so agents only spend time watching strangers.

Can AI recognize other objects?

Most camera AI is limited to generic objects, like vehicles, animals, humans and packages. As these algorithms continue to evolve, they are getting additional capabilities. One of the best examples is Arlo’s beta program to train cameras to recognize specific objects or states (like if a garage door was left open), so the camera can send an alert if it sees a certain thing or if a particular item disappears.

Do you have to pay for AI features?

That depends on the security company behind them. Brands such as Eufy, Lorex and Google Nest usually offer basic object recognition for free. Nest’s version, which comes with package detection, is particularly accurate. Other brands, like Arlo, Ring and Blink usually require a subscription fee to unlock AI detection and related smart alerts, starting between $3 and $8 a month.

When you reach more advanced features like face recognition or customized object detection, a subscription is required as these options are locked behind higher tiers.

Do these AI features compromise my privacy?

Generic object recognition offers few privacy issues. However, if you want to share home security videos online or upload them to a platform, the camera brand may receive permission to use your video to help train their AI.

Facial recognition features are more complicated. Face data must be shared with security companies when creating face profiles. Companies typically use a “not-quite-end-to-end” encryption system for face profiles, which only decrypts the data when it’s needed. If someone else (like a homeowner you’re making deliveries to) creates a face profile for you, you don’t really have control over what happens to that profile.

Privacy concerns like these are also being worked out via law. Visit Illinois, for example, and all these face recognition features are disabled because of the state’s current privacy laws. Other cities and states are working on similar legislation.

Are these AI features available in business cameras, too?

They often are, offering a way to cut down on unnecessary alerts and even false alarms. Security companies are also working on additional AI recognition capabilities with commercial applications. That includes training models to recognize suspicious behaviors, such as movement associated with theft or lurking.

Do cameras have any other AI tricks?

They will. It’s been a huge year for security camera AI, but the growth of these algorithms continues. One example is Swann Security’s conversational AI for video doorbells, which can answer questions about who is living at the residence for the homeowner (as rudely or nicely as you’d like). Expect home security camera AI to offer even more tailored options and start acting increasingly like your personal bouncer or butler in the coming years.

To learn more, check out how a home security system can help save you money, the worst spots to put a home security system oath how to find professional home security monitoring for cheap prices.