Tell HN: Meta's AI support feature allows Instagram accounts to be stolen
16 points
6 hours ago
| 4 comments
| HN
If the AI support option is enabled for your Instagram account (it appears to be A/B tested for only a percentage of accounts), anyone can hijack it with little effort. Simply get on a proxy or VPN close to the account's region, then ask the agent to send a code to an arbitrary email address. Once you receive the code, pass it forward to the agent, and it'll provide you with a password reset link which you can then use to sign into the account.

Posting here for any Meta employees who may be reading. This flaw has been around for at least a few days and has been used to hijack over 100 high-value Instagram accounts. The correct patch would be to disable the AI support feature entirely for the time being until this is sorted and revert accounts and usernames that have been hijacked over the last few days. This is a pretty important flaw and it's currently being exploited in blackhat circles. The steps above are public knowledge in these circles and can be found trivially on Telegram.

Edit: I wouldn't be surprised if this was never acknowledged by Meta. Several months ago in February, there was an exploit that allowed anyone to view the email address and phone number on file for any Instagram account. No acknowledgement from Meta. IMO they should've filed an SEC 8-K for an issue like that. Also, this flaw was unpatched when I posted about it - not sure if it's since been patched.

LocalExt
6 hours ago
[-]
I think AI for social channels are really unnecessary for users because of 80% hallucination. I know these AIs only useful for channels owners to track and learn user's habits and collect data...and data is money.
reply
rolls
3 hours ago
[-]
This happened to my account today. My sessions were revoked and password changed with no email, text, or push notification. Email and text codes weren’t being sent to my phone. I went through several cycles of resetting my password then getting hijacked again.
reply
hcwilk
3 hours ago
[-]
I didn't see it in the original post, but is there any way to turn this off at an account level?
reply
king_zee
5 hours ago
[-]
When will people learn granting any kind of account access to an LLM is a moot point, if the LLM has knowledge of something, by design it can't help but divulge it
reply
jmye
28 minutes ago
[-]
> No acknowledgement from Meta.

Of course not. Every single employee is a sociopathic as their CEO.

reply