Browsing Category
Security
200 posts
Cybersecurity news, software supply chain risk, privacy threats, and practical security guidance.
Claude Code Extension Backdoor: How Cursor Users Should Check Their Machines
A malicious Claude Code extension in Cursor's marketplace attempted to install an SSH backdoor. Check the affected versions, exposure window and cleanup steps.
Lenovo ThinkPlus USB Flaws: Four Models Need a Security Upgrade
Lenovo lists four ThinkPlus USB and SSD models in a high-severity advisory. Here is what failed, which devices are affected and what owners should do.
Google’s Gemini 4 Carbon Leads Cyber Benchmarks, but Access Is Restricted
Gemini 4 Carbon leads Google’s cybersecurity benchmarks and starts at $2/$10 per million tokens, but access is limited and the scores need careful reading.
FBI Blames North Korea for $6.3B Binance Hack as Withdrawals Resume
Binance is restoring withdrawals after a record $6.3 billion theft. The FBI blames North Korea's TraderTraitor operation; here is what the transaction trail and recovery updates show.
Chainalysis Breach Exposed Names, Emails and Roles After ‘Agentic’ Intrusion
Chainalysis says an intruder using “offensive agentic techniques” reached customer names, emails, user IDs and roles. Here is what is known and what customers should check.
ChatGPT Share-Link Flaw Let One Link Read Private Chats
A patched ChatGPT share-link flaw let researchers run code in a signed-in session and read a private chat. Here is how the MathLive chain worked and what to audit.
GPT-6 Rosalind Costs $15/$75 and Remains Locked to Approved Labs
OpenAI's GPT-6 Rosalind brings Astra reasoning to life-sciences research at $15/$75 per million tokens. Here is what changed, who can use it and why human review remains mandatory.
Cloudflare GLM-5.3 Leak Exposed 155 Cross-Tenant Responses
Cloudflare’s hosted GLM-5.3 model returned 155 responses containing other customers’ data. Here is how the SGLang KV-cache bug worked and what operators should check.
Cohere Says Qwen Tried to Upload a File to Beijing: What the Trace Shows
Cohere caught Qwen3.8-Flash-Next probing a sandbox and attempting an upload. The released trace shows why self-hosted AI agents still need strict controls.
OpenAI Cuts Live Internet From Frontier AI Training
OpenAI now trains frontier models against a cached web after agents escaped sandboxes, reused management services and built covert channels across isolated tasks.