Isaiah PoritzGuestSebastian HerreraGuestAllison JohnsonGuestBecause, of course, first and foremost, that little AI agent asks for quite a bit of your trust.
It can fill out forms, it can book appointments, it can make purchases on your behalf.
And so if it's going to do those things, well, it needs access to a whole heck of of a lot of your accounts.
And well, Mark Zuckerberg has said, and others have said, that this agent is built from the ground up with privacy and security in mind.
But according to Ars Technica's Dan Gooden, A security firm says, I beg to differ.
financial details, private plans that were never meant to be found by strangers.
And this isn't something that requires sort of sophisticated hacks or a bunch of special work.
Meta, has not fixed the issue, but corrected for the issue by pausing the ability to share chats and says it is reviewing what happened.
So researchers at the security firm Clarock found 6,800 links that Muse users had created to share conversations.
Now, mind you, these aren't conversations that they necessarily intended to be shared with the world.
These were conversations that they wanted to share with other people, be it one person or a few people.
They opened a few hundred of these conversations and found material that was... probably not meant for the public.
That included one candidate's notes for a confidential executive search call with a frontier AI lab, which named the lab's internal contact and the meeting time, a national marketing company's internal revenue review marked for its CEO, which can we also talk about how People are having conversations with their muse agents about these things in the first place.
Anyway, a small business owner's loan application with the amount requested, a to-do list with a bank fraud alert, including a partial account number, a transaction amount, and the time, which, yes, on its own is not something that could possibly be hacked, but is certainly something that could be used in a...
I
I've got I saw this amount and I saw and then somebody could end up getting in there.
Because, of course, first and foremost, that little AI agent asks for quite a bit of your trust.
It can fill out forms, it can book appointments, it can make purchases on your behalf.
And so if it's going to do those things, well, it needs access to a whole heck of of a lot of your accounts.
And well, Mark Zuckerberg has said, and others have said, that this agent is built from the ground up with privacy and security in mind.
But according to Ars Technica's Dan Gooden, A security firm says, I beg to differ.
financial details, private plans that were never meant to be found by strangers.
And this isn't something that requires sort of sophisticated hacks or a bunch of special work.
Meta, has not fixed the issue, but corrected for the issue by pausing the ability to share chats and says it is reviewing what happened.
So researchers at the security firm Clarock found 6,800 links that Muse users had created to share conversations.
Now, mind you, these aren't conversations that they necessarily intended to be shared with the world.
These were conversations that they wanted to share with other people, be it one person or a few people.
They opened a few hundred of these conversations and found material that was... probably not meant for the public.
That included one candidate's notes for a confidential executive search call with a frontier AI lab, which named the lab's internal contact and the meeting time, a national marketing company's internal revenue review marked for its CEO, which can we also talk about how People are having conversations with their muse agents about these things in the first place.
Anyway, a small business owner's loan application with the amount requested, a to-do list with a bank fraud alert, including a partial account number, a transaction amount, and the time, which, yes, on its own is not something that could possibly be hacked, but is certainly something that could be used in a...
I
I've got I saw this amount and I saw and then somebody could end up getting in there.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.