ChatGPT Users Data Leaked
twitter.com
twitter.com
A bug in some code that we took for free and didn't review.
2. how do you know they did not review it? the problem obviously presented in an edge case, which is by definition hard to catch.
3. they helped find and fix a bug in the library, which is one of the ways of giving back to the community.
So you're telling me that you've reviewed every line of every library you've ever used?
Also, your saying doesn't mention anything about people who point out to others who should be stoned. I'm assuming that's okay, as well as letting people know what they did that would deserve a stoning. Just in general, not advocating anyone be stoned.
If it's something completely unpredictable then, no, they're not generally liable. It's called an "act of god" and you'll find that language in virtually every contract.
> Also, your saying doesn't mention anything about people who point out to others who should be stoned. I'm assuming that's okay, as well as letting people know what they did that would deserve a stoning. Just in general, not advocating anyone be stoned.
I have absolutely no idea what your point is here.
Parent wasn't lobbing stones, just pointing out that we are responsible, if not culpable, for the results of the development decisions we approve and commit. A library was used that wasn't fully reviewed, vetted, or corrected prior to use. It's like buying rivets and not inspecting them before you build a skyscraper with them - even if nothing comes of it, was no wrong done?
There was talk somewhere about a bug bounty program being a good idea. I have to agree.
It turns out that if you put control signals in-band in the data channel, it's extremely hard to get right, and can easily lead to issues like this if connections aren't managed very carefully. One could make a comparison to how LLMs work at the moment too, it feels like a related class of vulnerability to me.