verge-ai · OCT 1
SOURCES: AMAZON SEEKS TO OFFLOAD $8B OF NVIDIA CHIPS TO INVESTORS... THEN LEASE THEM BACK
techmeme · OCT 2
Tech executives sign voluntary AI safety agreement...
wired-ai · OCT 1
AI TOOL RECONSTRUCTS IMAGES YOU'RE LOOKING AT... FROM YOUR BRAIN SCAN
mit-tech-review · OCT 1
STUDY: One 'verified source' note flips 45-88% of correct AI answers in 7 of 8 models
r-machinelearning · OCT 1
REPORT: OpenAI cuts ties with 3 safety researchers over mishandled sensitive info
techcrunch-ai · OCT 1
Flaw in ChatGPT's Mac app could have let hackers grab sensitive data
wired-ai · OCT 2
Nvidia chips keep reaching China despite US curbs... Officials ask why company missed red flags
techmeme · OCT 2
OPENAI UNVEILS 'DOTS' AI AGENT... BUT ONLY FOR $100/MONTH PRO USERS AND UP
verge-ai · OCT 1
GOOGLE REVEALS GEMINI 4 ARGON... SO CAPABLE ONLY 'TRUSTED CYBER DEFENDERS' CAN HAVE IT RIGHT NOW
r-machinelearning · OCT 1
SOURCES: Clayton likely White House pick for AI czar... may stay as intel chief
techmeme · OCT 2
Judge tosses antitrust suits over Google AI Overviews
verge-ai · OCT 1
Google launches first advanced chip into orbit to pave the way for space data centers
techcrunch-ai · OCT 1
[PAPER] STUDY: HIDE A SECRET IN A URL, AI FETCHES THE PAGE... ATTACKER GETS THE DATA
arxiv · OCT 1
[PAPER] STUDY: Chained 'skills' hijack AI agents in 74% of attempts
arxiv · OCT 1
[PAPER] STUDY: New defense blocks AI agent attacks at the last second before a tool call runs
arxiv · OCT 1
arxiv · OCT 1
[PAPER] STUDY: New training method beats all 7 multi-turn jailbreak attacks on 5 open models
arxiv · OCT 1
[PAPER] STUDY: Safe-response 'prototype' score inverted on jury-labeled control... ranks unsafe above safe
arxiv · OCT 1
[PAPER] Paper proposes open-source AI Assessment Sandbox Configurator for EU regulatory sandboxes
arxiv · OCT 1
arxiv · OCT 1
[PAPER] STUDY: Fine-tuned Qwen2.5-7B shifts ER triage level in 5.27% of bias tests, vs 16.02% for base model
arxiv · OCT 1
[PAPER] STUDY: Training-free AI screen could take over a third of water-system alarm reviews
arxiv · OCT 1
[PAPER] STUDY: Every AI agent can be right and the team still wrong... missing shared state, not smarts
arxiv · OCT 1
[PAPER] YouRA: New research agent keeps persistent state to trace paper claims back to executed experiments
arxiv · OCT 1
[PAPER] STUDY: Visual corruption hits self-driving AI models unevenly — and a fix helps only some of them...
arxiv · OCT 1