What Is AI Telling Prospective Employers About You?

**SPEAKER_1** (0:00)
This episode is sponsored by HPE. Today's networks must evolve from simply providing connectivity to harnessing AI to deliver exceptional connected experiences. At the break, HPE's Rami Rahim shares how self-driving networks can make that happen.

**Imani Moise** (0:19)
Welcome to Tech News Briefing. It's Friday, July 17th. I'm Imani Moise for The Wall Street Journal. Security experts say not to give your passwords to a chatbot, but AI agents can't do much without access to your accounts. We're test driving a new feature from a popular password manager that lets Anthropics Clawed login securely. But should you hand over your credentials? Then, routine background checks are starting to dig deeper with the help of AI. We'll explain how technology is changing the way employers vet job candidates, and why you shouldn't be surprised if your bachelor party photos come up in a job interview.
It all starts with a secret handshake. A new integration from password manager OnePassword lets Clawed AI agents access your accounts without revealing your passwords. Here's how it works. When Clawed needs to log in to one of your online accounts, it asks OnePassword to handle the sign-in. After a biometric authentication, OnePassword fills in your credentials without exposing your password to the chatbot or Anthropics servers.
But just because you can grant access to your accounts doesn't necessarily mean you should. WSJ personal tech columnist Nicole Nguyen put this new feature to the test and is here to explain how it went, and which accounts she would think twice about connecting.
Nicole, how did you feel in the moment that you gave Clawed access to your passwords?

**Nicole Nguyen** (1:44)
Very nervous. I think I was, I felt some sweat forming, you know, because I've been covering hacks for a long time. And also I've been testing AI agents for over a year now, and I've seen how they do things that are unpredictable. The CTO of one password who I talked to, Nancy Wang, said that agents act like humans, but they work like machines. And by that she means they're making decisions like a human, but like a machine, it can go on forever. We clock out at the end of the day, but an agent does not.
And so it is persistent, and if it's not satisfied with how it's finished the task, it will keep going and going and going, and sometimes that has very unintended consequences.

**Imani Moise** (2:30)
So which task did sharing access allow you to offload? And what was the process like?

**Nicole Nguyen** (2:35)
So I started with some low-stakes stuff because I was afraid of giving Claude access. I knew that one password wasn't gonna share my password with Claude, but Claude still got into my accounts and could see into those accounts and could take action in those accounts. So I wanted to be careful. So first I just had to do simple stuff, like renew my library books. Okay, simple enough. Then I had it log in to wsj.com and find our staff's June fiction book recommendations and add those to my read later list, which it did totally fine. So I was like, okay, let's ratchet it up a little. I found a flight I liked on Google flights and had Claude log in to my Chase account to see how much that flight would cost in credit card points and turns out credit card points weren't worth it.
That's something that I would have done anyway, but Claude spared me the 10 minutes or whatever of trying to navigate Chase's travel portal.
It did okay. Then for the purposes of content, I wanted to see with a sensitive account how Claude would behave. So I gave it access to my credentials and one password to my 401k account. I don't recommend you give an AI agent access to your life savings, but it did tell me that my investment election for my 401k was costing me over $400 annually in fees and proposed a new fund mix that would cost me $15 a year. I haven't taken action on that yet, but I thought like this is pretty useful and I'm glad that I gave it access one time, but I think ultimately there are some jobs that are better left to humans.

**Imani Moise** (4:23)
It sounds like you're probably least comfortable sharing access to your 401k account, but are there steps that people should take before connecting more sensitive accounts like healthcare portals or bank accounts to these services?

**Nicole Nguyen** (4:35)
At this point, I wouldn't necessarily give an agent access to those sensitive accounts because sending money to the wrong person or changing your insurance type could have catastrophic consequences.
Think of an agent in your personal life as a young intern slash assistant who can take care of the menial, mundane stuff that you hate to do and leave the high-stakes stuff to yourself so that you have more time for that more meaningful or more sensitive work. So I'm talking things like online grocery shopping, or if you like to review your utilities bill every month to make sure your energy or water usage isn't spiking, let an agent do that instead of things like banking, or investing, or messaging your doctor.