AI & LLM
AI and LLM covers generative models, agents, prompt injection, data security and new artificial intelligence tools. The collection connects capabilities, limitations, operational risks and their impact on technical work.
Autonomous AI Delivers Full Ransomware Attack in 10 Hours — and an 80-Page Audit for the Victim
Palo Alto Networks' Unit 42 has documented the first ransomware attack entirely managed by AI agents, completed in under 10 hours. The…
OpenAI's Astra Crosses the Critical Threshold: AI That Finds Zero-Days Without Guidance
OpenAI has designated Astra as the first model to reach the Critical threshold of its Preparedness Framework. The system discovers zer…
AI AppSec: Scanners Agree on Just 5%, Triage Costs $128,000
Contrast Security's AppSec Overflow 2026 report reveals three AI scanners testing the same codebase agree on only 5% of findings, whil…
Anthropic: Infostealer Malware Steals Claude Sessions, Drains Usage Credits
Anthropic has warned Claude users that infostealer malware on compromised endpoints has stolen active sessions. Attackers are using th…
InstallFix Campaign Clones Claude Code to Spread Malware Without Exploits
Cybercriminals are using malvertising and cloned sites to distribute stealers and backdoors. The attack exploits no software vulnerabi…
LLM Safety Rests on 50 Neurons: How Unit 42 Shatters the Myth of Distributed Alignment
Unit 42 research shows that safety mechanisms in large language models reside in less than 0.02% of neurons. The perturbation probing…
AI Kill Switch Act: The Red Button Is Law, But Does It Work?
The bipartisan AI Kill Switch Act requires developers to maintain the ability to shut down advanced AI systems. The distributed nature…
Cisco Challenges National Labels on AI Models: Nearly 900 Fingerprinted
Cisco and VAIL research reveals 'provenance entanglement' — upstream weights, biases, and behaviors cross borders without appearing in…
Aurora Ransomware Group Abuses AI Cursor Agent for Post-Compromise Attacks
The Aurora ransomware group used the AI-powered Cursor Agent with Claude Sonnet as an interactive operational assistant during active…
AI Honeypot: Real Attacks on LiteLLM and MCP Servers Reveal Three Patterns
Wiz Threat Research deployed AI/ML honeypots for 90 days and documented sustained, service-specific attacks. Three distinct patterns t…
NSA and CISA Issue First Alert on AI-Driven Attacks Against Critical Siemens PLCs
Five U.S. federal agencies have released joint advisory AA26-231A confirming threat actors are using AI-generated scripts to target in…
Exploitarium: The 'Recruitment by Chaos' That Shatters the CVD Model
Pseudonymous researcher 'bikini' dumped over 30 zero-day PoC exploits on GitHub on June 27, 2026, without any vendor coordination. CVE…