GROK
Another week, another AI privacy mess
Hundreds of thousands of conversations with Elon Musk’s AI chatbot, Grok, have been turning up in search engine results, and many users likely didn’t know they were public.
The problem comes from Grok’s “share” button, which creates a unique link when users send a chat transcript to someone else.
But instead of keeping those links private, Google indexed almost 300,000 of them, making the chats searchable for anyone.
Some of the indexed transcripts show users asking Grok to create secure passwords, draft weight-loss meal plans, provide medical advice, and even push its limits.
In one example seen by the BBC, the chatbot gave detailed instructions on making a Class A drug.
Grok isn’t alone here.
OpenAI faced a similar issue earlier this year when shared ChatGPT conversations started appearing in search results, leading the company to scale back its experiment.
TL;DR
Around 300,000 Grok chats have been made searchable on Google.
Shared transcripts can still reveal personal or sensitive information.
OpenAI and Meta have both faced similar privacy concerns before.
Oops, all chats
Meta also came under criticism after chats with its Meta AI chatbot were made visible in a public “discover” feed.
Experts warn that even anonymised transcripts can still expose sensitive personal details, like full names, locations, health information, or business plans.
Once indexed online, these conversations can remain accessible indefinitely.
Hey guys? A friendly reminder: AI receipts are forever.