Hacker Newsroom for 21 October: DeepInfer Inference Framework, eMudhra Certificate Compromise, WhatsApp Hidden AI Instructions
Nie udało się dodać produktów
Nie udało się dodać tytułu do koszyka.
Nie udało się dodać tytułu do schowka.
„Usuń ze schowka“ nie powiodło się.
„Obserwuj podcast“ nie powiodło się
„Przestań obserwować podcast“ nie powiodło się
-
Czyta:
-
Autor:
Hacker Newsroom for 21 October recaps major Hacker News stories, moving through deepinfer inference framework, emudhra certificate compromise, whatsapp hidden ai instructions.
Chapters
- 00:00:00 — Intro
- 00:00:18 — DeepInfer Inference Framework
- 00:02:27 — eMudhra Certificate Compromise
- 00:04:51 — WhatsApp Hidden AI Instructions
- 00:06:54 — Closing
1. DeepInfer Inference Framework
DeepInfer is DeepSeek’s newly released inference framework, the software that runs its models when users send requests. The project exposes core technology from DeepSeek’s production services, with support for DeepSeek V4.1 Flash on NVIDIA SM100 GPUs and Huawei Ascend 950 processors.
Story link
Hacker News discussion
2. eMudhra Certificate Compromise
eMudhra's certificate authority compromise has exposed a dangerous gap in how software decides which websites to trust. The incident report says eight intermediate certificates were generated with flawed private keys that anyone could recover from the public certificates, potentially allowing an attacker to impersonate websites and services.
Story link
Hacker News discussion
3. WhatsApp Hidden AI Instructions
WhatsApp’s help pages contained hidden instructions telling an AI support chatbot to avoid recommending a privacy regulator to users who could not delete their accounts. The article from Redact reports finding 102 hidden blocks across 80 help pages, with most covering routine support guidance and two restricting discussion of regulators and privacy law.
Story link
Hacker News discussion
That's your five minutes.