I saw a headline that Muse was leaking people's contact info. There was a DEFCON presentation in 2025 on how they (hackers) coerced private/personal data out of a model (openai maybe). That has been patched most likely.
Meta doesn't have a great track record of protecting data.
I don’t think it’d be related to this if the defcon presentation was in 2025, but your description reminds me of this HN post about getting Claude to leak memories: https://news.ycombinator.com/item?id=48916975
A few years ago, I had a client who sold personalized spicy stories. That is before AI, there was a simple web order form, and they would write based on whatever their client requested.
We put ads on Google and Google suggested to add stock photos from its library. So created a few ads using those photos. Nothing crazy, just muscular men without shirts.
Next morning we woke up with our Google Ads account being suspended due to repeated violations because we had multiple ads.
Tell your muse to talk to their support muse to resolve this
This made me laugh harder than it should have.
Agents all the way down.
I saw a headline that Muse was leaking people's contact info. There was a DEFCON presentation in 2025 on how they (hackers) coerced private/personal data out of a model (openai maybe). That has been patched most likely.
Meta doesn't have a great track record of protecting data.
But their record of failing to protect data is rock-solid!
I don’t think it’d be related to this if the defcon presentation was in 2025, but your description reminds me of this HN post about getting Claude to leak memories: https://news.ycombinator.com/item?id=48916975
As a German speaker, I read this and found it sehr amüsant.
A few years ago, I had a client who sold personalized spicy stories. That is before AI, there was a simple web order form, and they would write based on whatever their client requested.
We put ads on Google and Google suggested to add stock photos from its library. So created a few ads using those photos. Nothing crazy, just muscular men without shirts.
Next morning we woke up with our Google Ads account being suspended due to repeated violations because we had multiple ads.
AI will actually encourage large scale breach.
By doing so, the data is effectively public now so later can be trained and they can still deny direct use.
The marketing makes it sound like it can do anything. Then you actually use it and discover how little it can reliably do.
Isn’t that a good thing if it’s actually related to AI usage?
It is a bit dumb tho if they are going to ban people using their AI they should just instruct it to not publish to their own products.
Ask muse to appeal with support agent.
Meta is like they can scrape and do everything with other people website, but others aren't allowed to scrape their content.
Meta moderation AI is very aggressive. It is not uncommon for people to get banned for using their own tools like the whatsapp API.
Do you have any idea what Muse created that set off the moderation?
They don't tell you much but I'm guessing that it fixed typos, spelling mistakes and reposted too many ads instantly which violated some usage rule.