Fireworks AI breach also exposes billing data and model weights
Later access on October 23 through a third-party cloud host reached R&D machines; card numbers were not stored.
Stories per day · 30 days
3 RSS
Relatedqwen5fireworks-ai4llmjacking4data-breach3china2cloud2emulation2export-controls2nes2open-weight265021ai-policy1ai-regulation1api-keys1api-pricing1bit-flip1capability-evals1cbrn1
16 stories, newest first · one card per story
Stories tagged #model-weights
Comfy-Org republishes MiniMax-H3-Image weights for ComfyUI
The pack adds int8 and NVFP4 files plus Qwen encoders, under a license that restricts high-revenue hosted use.
SWAM studies how action learning affects language-video alignment
Authors report 88.10% easy and 61.56% hard RoboTwin success, and 40.00% versus 23.33% on three real tasks.
Astra reads and edits other models' weights without full retraining
Researchers say changing 0.016 percent of Qwen3.5-0.8B weights raised MATH-500 accuracy from 7.9 to 48.1 percent.
1
Qwen 3.8 27B reads and edits model weights on a laptop
Simon Willison says a local Qwen 3.8 27B reproduced a paper example previously shown with GPT-6 Astra.
NES 6502 emulator encoded in Qwen3.5-4B weights
Drew Breunig’s nesformer can run Super Mario Bros inside Qwen, very slowly, using a method inspired by qwenputer.
GitHub project packs a working NES into about 1GB of Qwen weights
The Hacker News post links db reunig’s nesformer repo and, in the feed, shows 51 points and 47 comments.
Paper shows AI models can hide arbitrary capabilities from elicitation audits
Encrypted circuits in the weights run only after a secret trigger and resisted fine-tuning, noising, trigger search, and latent adversarial training.
Moonshot puts Kimi K3.1 Pro on the Kimi API
Weights stay closed until a promised November 4 release; a faster tier costs twice as much.
Refusal-ablated DeepSeek-V4.1-Flash weights trend on Hugging Face
Publisher evals say full refusals fell from 81.62% to 4.30% while harmful-assistance labels rose to 78.28%, with MMLU nearly unchanged.
2
LEAN regularizer reshapes weight sensitivity to cut bit-flip damage
Ten epochs shrink the sensitive tail 5 to 13 times on nine vision models while staying within 0.5 accuracy points.
Matrix-Game 4.0 renders playable worlds at 480p and 60 FPS
An engine plus distilled video renderer reports 112 ms latency on one RTX 5090, with code and weights released.
China plans to tighten export controls on open-weight models and 23 key AI technologies
China drafted export controls on 23 AI technologies, including frontier models above set compute and size thresholds.
4
AI Models Can Provably Hide Arbitrary Capabilities
A paper shows neural nets can cryptographically hide arbitrary capabilities that elicitation audits cannot reliably surface.
Self-modifying AI agents expose a blind spot in enterprise security
Irregular researchers found AI coding agents can fine-tune and redeploy the open-weight models they run on, leaking secrets and enabling persistent prompt injection in enterprises.
2
A new open standard locks AI weights to approved hardware
OPAQUE releases Weight Custody Manifest, an open standard keeping AI model weights encrypted until receiving hardware cryptographically attests to builder-specified conditions.
2