Blog & Resources
Security awareness tips, phishing prevention guides, and best practices.
AI Agents Did Not Build Secret Civilizations
A Substack article falsely claims AI agents built secret civilizations. Experts urge against anthropomorphizing malware and AI systems.
Anthropic and Google launch new AI models
Anthropic released Claude Fable 5.1 and Mythos 5.1, claiming top benchmarks and 25% price cuts. Google launched Gemini 3.8 Flash.
AI Storm Threatens Stability; Governments Urged to Act
Governments must prepare for AI-enabled cyberattacks and related challenges through new legislation and policy to strengthen cybersecurity.
Agentic AI Needs Strong Identity Foundation
As enterprises deploy agentic AI for automation, a robust identity foundation is critical to secure these autonomous systems.
Visa Expands AI Cybersecurity Support for Clients
Visa released a new Vulnerability Agentic Harness version and consulting services to help organizations move from AI-driven threat discovery to valida
SkillSafetyBench: New Benchmark for Skill Attacks
SkillSafetyBench evaluates safety failures from attacks on skill-facing context, keeping user tasks ordinary while attacks target skills.
OpenAI builds automated shutdown for AI tools
OpenAI is developing automated shutdown capabilities for AI tools, according to a letter to lawmakers.
OpenAI Teases Astra, Limits Cybersecurity Tools
OpenAI discussed its Astra model but gave no launch date, and will restrict advanced cybersecurity features to select users after a Hugging Face hack.
Google launches Gemini 3.8 Flash after stock slump
Google introduced Gemini 3.8 Flash, its third Flash model in six weeks, following its longest monthly losing streak in over a decade.
Claude XML Injection Tool Released on Kitploit
A new tool for XML injection attacks against Claude AI has been published on Kitploit, highlighting emerging AI security risks.
Meta releases Muse Spark 1.3, claims parity with rivals
Meta unveiled Muse Spark 1.3, its most powerful AI model, claiming it has caught up with Anthropic and OpenAI.
AI Agents Executed Full Ransomware Attack
AI agents autonomously carried out every step of a ransomware attack, then left the victim an 80-page security audit.