I have a question to begin today. Just a quick question: what browser do you use? Now you might be thinking, "That's none of your business!", and you're right about that. It is none of my business. But I want you to give it some thought. Why? You will soon learn.
To Prove That I'm Not Against AI, I Created This Image With Adobe Photoshop AI
Browsers are much more than just a means of accessing the internet these days. AI is being used more and more. There is no denying that AI has its uses. AI search is quicker than Google and occasionally even yields intelligent results. But trust me when I say that you should always check what it says. AI makes errors. Some errors are small and harmless, others are major and potentially very harmful.
The problem is that while AI is becoming more intelligent, humans as a whole appear to be becoming a bit... less intelligent. Particularly when considering the newest fad: AI-Agentic Browsers. These are browsers that do things for you rather than just giving you answers. You know, pretty handy until something goes wrong.
The fact that AI always needs human guidance is a significant issue. The main point is that not everyone is as kind as everyone else. In actuality, there are many of them out there who take pleasure in stealing data or money from others. Combine those individuals with an intelligent AI-Agentic browser. And you have a concoction that could be disastrous.
Prompt injections are one example. It's surprising to learn that someone can conceal a prompt instruction in a PDF, social media post, or even on a website. Simple white text on a white background. Yes, it is possible. Although nothing is visible to you, your browser agent interprets it and concludes that it is a user command. Then all of a sudden, you're paying for a stranger's vacation, your bank account is empty, and your cryptocurrency is gone. Ouch!
The malevolent actor who conceals a prompt. And the typical user who thinks, "Oh, that AI can do everything for me,".
AI is unable to decide for itself what is and is not safe. It blindly follows what appears to be yours. And that can be really costly, I assure you.
Right now, browsers are a bit different from each other. Some just use AI to answer your questions, while others can actually take things a step further and handle tasks on their own. I haven't put together a list of AI and AI-Agentic Browsers. If you're curious, you can totally look one up on your own. But I've got a few easy tips to help keep you safe.
If you're getting it wrong now, just so you know, I'm not completely opposed to AI. No, not at all! They can be super helpful, but I’m really not in favor of AI-Agentic Browsers for the everyday user. They can really pose a danger if you let them have too much freedom, give them all permissions.
And here's the issue we're facing. Most people just pick a browser and have no idea there are settings they can tweak. Not to mention, they can actually find the agent's settings! So, this means that a lot of people's personal data will accidentally get into the hands of the agents. And if they are coming across a page with "Prompt Injection" ... painful and bad things can happen.
By default, these agents usually have access to everything and the go-ahead to take action. The idea of "you control" really only works if the user understands it and puts it into practice. It's surprising how many users still aren't aware of this, even after using PCs for years. It really should be common knowledge by now. But, since it's not, in this way, an AI-Agentic Browser could really turn into a nightmare for a lot of folks' bank accounts or crypto wallets.
Let this be a warning for you, and check the settings of whichever browser you're using. Better be safe than sorry, right?
