Browsing Tag
AI Cybersecurity
18 posts
Cybersecurity risks, defenses, and policy tied to artificial intelligence systems.
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.
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.
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.
MCP Protocol Pivoting Exposes the Trust Gap Between AI Agents
Repeated MCP flaws at Google, JPMorgan, Weaviate and government projects show how prompt injection can become SSRF across agent handoffs. Here is what builders should fix.
Gemini 4 Argon Debuts With 1M-Token Output, but Access Starts With Cyber Defenders
Google's Gemini 4 Argon pairs a one-million-token output limit with low introductory pricing, but only vetted cyber defenders can use it first.
Palo Alto Launches Always-On AI Pentesting With Mythos 5 and GPT-5.6-Cyber
Palo Alto Networks' Unit 42 now uses Mythos 5, GPT-5.6-Cyber, and open models for continuous attack-path testing. Its own results show why buyers should judge the harness, controls, and remediation workflow, not just the models.
Cisco’s Twice-Monthly Patch Cadence Starts With Catalyst Center and ClamAV Fixes
Cisco’s first July security-advisory drop under its new twice-monthly cadence includes a Catalyst Center arbitrary-file-read flaw and seven ClamAV vulnerabilities affecting Cisco Secure Endpoint. The change gives network and security teams more predictability, but it also means Cisco infrastructure patch planning needs to become a standing operating rhythm, not a quarterly scramble.
Apple’s Early Security Updates Show AI Is Shrinking Patch Windows
Apple pushed iOS 26.5.2, iPadOS 26.5.2, macOS Tahoe 26.5.2, and Safari 26.5.2 out before the broader 26.6 release cycle, citing AI-driven security concerns. The update is a practical reminder that patch timing now matters as much as patch content.
Open-Weight AI Cyber Gap Narrows to Months, AISI Finds
The UK AI Security Institute says GLM-5.2 and DeepSeek V4-Pro now trail leading closed AI models on cyber tasks by roughly four to seven months. For defenders, that shrinking gap turns open-weight model policy into an operational security issue, not a distant AI governance debate.