Hacker Newsnew | past | comments | ask | show | jobs | submit | Aeroi's commentslogin



Ah. So what they are saying is that their 500TB CSAM was removed but they recovered 400TB. They were using it to build hashes. Yeah I would not expect them to share that with the world. It would not surprise me if automation removed it because it is CSAM.

Perhaps they should store such things on hardened secured servers in NSA data-centers (yes, related) instead of Google servers unless the plan is to integrate the CSAM into Google. Otherwise generate the signatures on hardened infrastructure that Google employees do not have access to and then share the signatures with all the hosting providers. (Checksums, Microsoft PhotoDNA, etc...)


The hack occurred in 2023, 500tb of data was comprised relating to Jeffery Epstein with 100tb not recovered.

Did you mean gb vs tb? AFAIK the original archive was 300GB but if there is another 100 to 200 GB that should be shared.

This was labelled a dupe by HN but where is the original and why does it link to age-verify?


also the link does not actually go anywhere, it goes to the age-verify page directly and then once you click "Yes"... nothing happens

it worked for me, but it did require the age verification.

How? The URL is literally going to `/age-verify`.

also wondering why this was labeled dupe, when the original link is nowhere to be found??

approximately 500 terabytes (TB) of data may have been compromised or accessed during the intrusion. The document confirms the intrusion occurred and discusses the security lapses (internet connectivity) that enabled it.

Where is the document?

i posted the actual link above, i'm not sure why its rerouting from the justice.gov website.

https://www.justice.gov/epstein/files/DataSet%209/EFTA001735...


in hindsight, the smart thing to do would have been to accept the contracts, knowingly enshittify the request, and protect other bad actors like Elon and xAI from ruthlessly compromising our democracies.

the administration continues to poison and insert itself into all aspects of American society.

two hardest problems for a platform like this.

1. users and initial flywheel. 2. content moderation.


what if, instead you just placed whatever weight you wanted onto a flat unfolded piece of paper.

open ai naming is a meme at this point


benchmark and pricing made me realize how good kimi 2.5 is. im an opus 4.6 person but wow, its almost 5x cheaper.


($10/$37.50 per million input/output tokens) oof


Only if you go above 200k, which is a) standard with other model providers and b) intuitive as compute scales with context length.


only for a 1M context window, otherwise priced the same as Opus 4.5


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: