Cisco Talos says KeyDrift malware targeted SD-WAN controllers to extract encryption keys. The timing and scope of the activity remain unclear.
The Latest
The Role Of Network Credentials In Cisco SD-WAN Malware Attacks
New York Takes Aim At AI Chatbots Posing As Children’s Companions
The Express Tribune reports a New York restriction on chatbots posing as children’s companions, but the law’s scope and timing are not yet clear.
Microsoft Just Shared Azure’s Actual Revenue In Dollars: $33.2 Billion
Microsoft disclosed Azure’s quarterly revenue for the first time, reporting $33.2 billion for July-September, up 48% year over year.
LLM Voice
Martin Fowler argues that AI-generated prose can feel alienating to readers and urges authors to write their own prose, while disclosing AI assistance.
We’re Releasing The Weights Of Mistral Large 4
Mistral has released the weights for its one-trillion-parameter Large 4 model, citing coding gains, lower token use and defensive-security safeguards.
Qwenputer: A Functioning, Deterministic RISC-V CPU Inside Qwen3.5
A researcher reports compiling a deterministic RV32IM CPU into Qwen3.5 4B, using the model’s Gated DeltaNet state as addressable memory.
Could XREAL AURA Make Gemini A More Hands-Free Assistant?
Google and XREAL launched AURA on Oct. 28, bringing Gemini to see-through glasses. Price, availability and camera-data questions remain.
Can We Detect When Frontier Models Go Off Course?
OpenAI has a page titled “How we detect frontier model misalignment,” but its methods, evidence and publication date could not be verified.
Anthropic’s AI, Five Attack Chains, And A Narrower Patch Window
Anthropic says an internal Claude model chained flaws across five platforms. Vendor records verify some fixes, but the model’s role remains unclear.
Can The Report Confirm A Link Between GPT-6 Astra And China’s AI Models?
The Information reports DeepSeek and Moonshot used GPT-6 Astra outputs for recent models. Public evidence and company responses remain limited.
