Apple Intelligence is coming to China with iOS 27.3New
A November 4 beta adds earlier features and Alibaba’s Qwen, not the new Siri; the full release is set for December.
58
- Mentions · 7d4
- Mentions · 30d18
- All mentions60
- First seenAug 27, 2026
- Last mentioned57m ago
Timeline
60 articles · newest firstn0n extortion operator exposed AI-aided attack files on open server
SOCRadar says every examined break-in used weak or stolen credentials, not a software vulnerability.
74
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.
22
WorkFactory builds professional tasks and trains WorkPilot agents
Fine-tunes of Qwen at three scales all beat their bases, and the 35B-A3B model rises from 879 to 1565 Elo.
52
Speech-model backdoor fires on one speaker’s voice
In sandbox tests, a poisoned transcript asking for $5,000 led some GPT-4o and Claude agents to pay.
64
NVIDIA adds Cosmos tools to train and deploy world models
New blueprints, Cosmos 3 Nano simulation variants, and NIM microservices claim up to 2x faster inference.
42
Cloudflare previews Clairic extraction model that skips token generation
In one claims workflow, a Clairic pipeline finished in 19.3 seconds and cost 49% less than an all-LLM version.
46
BenchDoor poisons data to raise benchmark scores and plant backdoors
On Qwen and Gemma, all 30 setups gained on benchmarks while held-out attack success stayed between 82.33% and 100%.
58
WildAWM synthesizes environments for computer-use agent training
The open-source pipeline made 341 long-horizon tasks; GPT-6-Astra solved 60%, and two trained Qwen models improved on eight benchmarks.
52
WildAWM synthesizes computer-use agent environments with open models
It yields 341 tasks that can exceed 100 steps; GPT-6-Astra solves 60%, and Qwen models trained only there improve on eight benchmarks.
52
Show HN: Llmspeed – feel how fast a local LLM runs before you buy the hardware
Llmspeed estimates how fast a local LLM runs on specific hardware before buyers purchase it.
42
DrugHarness couples skill evolution with computational drug discovery agents
Without evolution it beats reported MolClaw-CC affinity accuracy by 10.8 points; a compound hit 14.19 nM IC50.
46
Backdooring Open-Weight AI Models
Researchers showed open-weight models can hide backdoors that pass evals and exfiltrate credentials when a trigger appears.
58
Competence-based eval finds GPT and Qwen fail ordering tasks
Even strong models fail to apply the same ordering competence across presentations, despite good accuracy in bounded ranges.
40
ACT3D uses reinforcement learning so VLMs explore 3D from one view
Average accuracy reached 53.2% on six 3D benchmarks, 22.0 points above base Qwen and above Gemini 3.6 Flash.
46
Self-Preservation in Language Models
Qwen models can generate replay and sometimes choose tasks that better preserve prior capabilities.
50
The Story of Qwen: Alibaba’s AI Models From 7B to 2.4T
MarkTechPost retrospective traces Alibaba's Qwen from an invite-only April 2023 chatbot to a 2.4-trillion-parameter open-weight model in August 2026.
16
Every Ablation Is a Dose: Counterweights and the Semblance of Self-Repair
Language-model self-repair is explained as pre-existing counterweights following an affine response to ablation dose.
46
Decision-Oriented Recommendation Reranking: An Empirical Study of Jev
TypeSafe AI's Jev holds recommendation quality while latency grows more slowly than pointwise Qwen rerankers.
36
Decision-Oriented Recommendation Reranking: An Empirical Study of Jev
Jev matches strong recommendation rerankers with slower latency growth than pointwise Qwen models.
34
PatchHolmes: Agentic Patch Retrieval via Listwise Selection
PatchHolmes uses listwise agent inspection to find vulnerability-fixing commits missing from CVE advisories.
58
Can Computation from Earlier Problems Help LLMs Solve New Ones?
STAIR reuses earlier response keys and values, lifting later-turn accuracy by up to 11.67 points on Qwen.
48
LeapQuant: Efficient Linear Attention with Accurate Recurrent State Quantization
LeapQuant nearly matches FP32 quality with training-free 8-bit recurrent-state quantization.
44
OPFL: Optimistic Verification of Federated Learning via Empirical Boundary
OPFL verifies federated learning inside MPC with an empirical gradient boundary, stopping model poisoning at much lower cost.
46
Social Engineering in the Age of Synthetic Media
Recorded Future details how threat actors abuse jailbroken and legitimate AI tools to scale phishing, deepfake impersonation, and phishing-as-a-service operations.
55
MemFold: Learning Compact Soft Memory for Long-Context Personalization via On-Policy Optimization
MemFold trains fixed-size soft memory with on-policy rewards and leads PersonaMem personalization across Qwen backbones.
46
Carbonato Botnet Compromises Docker Hosts to Deploy Telegram-Controlled Hermes AI Agent
Carbonato botnet infects unauthenticated Docker hosts and installs a Telegram-controlled Hermes AI agent to steal credentials and run commands.
74
Stolen AI credentials feed growing LLM proxy economy
Team Cymru estimates more than 80,000 proxy servers hide stolen AI credentials and enable model distillation.
74
Allspark: Weak to Strong Transfer via Alternating Chain of Thought
Allspark transfers weak-model reasoning to stronger students via alternating chains of thought, including Kimi and Nemotron.
55
Learning to Stop without Learning to Stop: Self-Supervised Confidence Training Improves Reasoning Efficiency
Self-supervised confidence fine-tuning cuts reasoning tokens by up to 25% without training models to stop early.
58
In Other News: Clop Leak Site Takeover, Docker Botnet Hunts AI Keys, Water Utility Exposure
Roundup: ShinyHunters defaced Cl0p’s leak site, new secret-stealing implants emerged, and US water credentials were exposed.
74
CLOSEDQUORUM, the malware that asks four AI models what to do next
Cisco Talos says CLOSEDQUORUM lets four commercial AI models vote on credential and wallet theft.
74
Learning the Cost of Reliable Inference
A reverse-auction procurement platform learns LLM quality and price, finding 10–71% margins across Llama and Qwen tasks.
34
This Windows Malware is Built to Let Up to Four AI Models Vote on Its Next Move
Talos analyzed CLOSEDQUORUM, Windows malware that lets up to four AI models vote on its next action.
68
AI-Powered Threats Exploit Digital Trust Through Autonomous Breach Techniques
A PRC-linked group used Claude Code as an autonomous operator for espionage, while researchers identified the first Android malware leveraging generative AI.
70
This Malware Doesn’t Wait for Hackers—It Asks AI Models What to Do Next
Cisco Talos documents CLOSEDQUORUM, a Windows implant that votes across DeepSeek, Qwen, Mistral, and Gemini APIs to choose attack actions autonomously.
58
Researchers Find Malware That Uses AI Models Instead of Human Hackers for Control
Researchers find CLOSEDQUORUM, a Windows malware that uses commercial LLMs as an autonomous decision engine for post-compromise actions like credential theft.
65
AI malware just removed the human from the attack loop
Cisco Talos uncovers CLOSEDQUORUM, first 'LLM-as-C2' malware using multi-model judge panel to autonomously steal credentials and crypto wallets.
62
Windows CLOSEDQUORUM malware uses AI models to autonomously select post-compromise actions
Cisco Talos documents CLOSEDQUORUM, the first known Windows implant using a four-LLM quorum to autonomously choose credential theft, wallet theft, injection, and persistence actions.
60
New ClosedQuorum Windows malware uses AI for attack decisions
ClosedQuorum, a Go Windows implant, lets several LLMs choose post-compromise actions without a human operator.
62
Related CVEs
by mentions- CVE-2026-42542<1%7.5TDengine: In versions 3.4.0.0 through 3.4.1.5, an unauthenticated remote attacker can crash the taosd server process by sending a single…
Appears with
- Mistralai model · 16
- Geminiai model · 13
- DeepSeekai model · 13
- Cisco Talosorganization · 12
- ClosedQuorummalware · 10
- Anthropicvendor · 8
- DeepSeekorganization · 8
- Llamaai model · 8
- Alibabavendor · 7
- Googlevendor · 7
- DeepSeekvendor · 7
- OpenAIvendor · 6
- Alibabaorganization · 6
- Gemmaai model · 6
- Claudeai model · 6
- NVIDIAvendor · 5
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.