Props to @elonmusk for making this one of their priorities when he let the sink in, and it was quickly resolved. An open question for the reader is "why was this ignored for almost a decade?".
Props to @elonmusk for making this one of their priorities when he let the sink in, and it was quickly resolved. An open question for the reader is "why was this ignored for almost a decade?".
What we discovered is that, since Musk's takeover, that Twitter's CSAM scanning failed and was not noticed by Twitter. They fixed it several weeks after we notified them, and then shut off our API access to prevent further research.
Really?(!), how's that? It's a direct, verbatim quote from the (your?) blog post.
But thanks for dropping by, could you clarify a few things?
>[...] since Musk's takeover, that Twitter's CSAM scanning failed and was not noticed by Twitter.
What you're saying is that:
1. It was working before Musk was there.
2. It stopped working after Musk took over.
Please answer this in a non-subjective way.
>They fixed it several weeks after we notified them
And, in your opinion, is that a good thing or a bad thing?
It's a bad thing that it stopped working. It's a very bad thing that they didn't detect that it had stopped working. It's a bad thing that it took weeks to fix it. It's a good thing (well, good at best, i'd call it "expected" or "the bare minimum") that they fixed it.
wow, now i think you're just trolling.
The direct quote was this:
>Twitter had an apparent and now resolved regression allowing CSAM to be posted to public profiles [...]
You then went on to make several statements/claims, such as:
1. "@elonmusk for making this one of their priorities when he let the sink in"
2. "it was quickly resolved"
3. An open question for the reader is "why was this ignored for almost a decade?".
Each one of those statements are incorrect readings of the report and none of them are justified by the quote you pulled from it.
On (1): There's no evidence that elon musk made this a priority. Given that it broke after he bought twitter and that they didn't detect that it broke and that it took weeks to fix it when they were informed of it, I think the evidence points to the exact opposite (i.e. it's not a priority).
On (2): weeks is not "quickly", and the "quickly" was your own thing, the report only says that it was resolved.
On (3): as you've been informed, this wasn't broken for almost a decade, it was broken after musk acquired twitter.
I think it is good that they fixed it. That is why we directly reached out to them to help address the issue before we published. I think it is bad that they cut off our API access and are threatening academic researchers with lawsuits.
Twitter problems with CSAM were widely known way before Musk took over [1], so,
>[...] PhotoDNA scanning was working
with due respect and all, I find that statement questionable.
>I think it is bad that they cut off our API access and are threatening academic researchers with lawsuits.
Yes, that truly sucks.
1: https://endsexualexploitation.org/wp-content/uploads/Doe-v-T...
6 days later.
"J-j-just trust me bro ;)"
Timeline is:
* Twitter DID have this problem.
* Musk buys Twitter.
* After a couple months, Twitter does not have this problem anymore.
Who do you think was responsible for that?
A. The old engineering team who was 90% gone on a (laughable) tantrum?
B. Ella Irwin, Head of Trust and Safety who had been there for years and resigned after Musk came in?
C. Yoel Roth, Head of Integrity and Safety who had been there for years and resigned after Musk came in?
D. Elon Musk
I'm no Musk fanboy but the answer is evident.
I've been there for 15 years, never saw a single image that didn't pass muster and I've used it very intensively. So, unless you actively went and looked for it I am assuming you would not encounter it. But given the size of Twitter and the fact that any company that hosts images is going to be targeted you can assume that to the degree that it was going on before it is still going on today, only worse because Twitter has far less staff now to take care of these things.
To portray the former Twitter staff as aiding and abetting whilst Elon rode in on his white horse and 'took care of things' is not in line with reality as I perceive it.
And there is some proof that things have gotten worse, not better:
https://9to5mac.com/2023/02/06/twitter-api/
(and many other articles besides).