Blog & Resources
Security awareness tips, phishing prevention guides, and best practices.
Tech employees and leaders call for global AI risk
Employees from Meta, Anthropic, OpenAI, and Google have launched 'Pacing the Frontier' to advocate for a US-backed global effort to manage risks assoc
AI Industry Leaders Call for International Safeguards on
Over 1,100 AI industry professionals are advocating for an international framework to slow the development of frontier AI models if safety risks excee
New CryptanalysisBench Benchmark Evaluates LLM Capabilities
Researchers introduced CryptanalysisBench to test LLM performance against cryptographic primitives.
New CryptanalysisBench Benchmark Evaluates LLM
Researchers introduced CryptanalysisBench to measure the ability of large language models to perform mathematical cryptanalysis.
Linux Foundation updates Model Context Protocol to improve
The Agentic AI Foundation released an updated Model Context Protocol (MCP) that shifts to a stateless architecture and introduces security enhancement
Anthropic's Claude AI discovers faster attacks on HAWK-256
Anthropic researchers used the Claude Mythos Preview model to identify a key-recovery attack on the HAWK-256 post-quantum scheme and accelerate attack
Microsoft and Wiz AI Agents Outperform Single Models in
Microsoft's MDASH and Wiz's Project Atlas AI agents achieved over 90% success rates on the CyberGym benchmark by utilizing multi-model architectures t
OpenAI Sandbox Escape Highlights Need for Traditional
A recent sandbox escape involving OpenAI's AI agents demonstrates that foundational security principles like access limitation and isolation remain cr
Researchers Propose Methods to Improve AI Safety via 'Black
Researchers are developing techniques to identify specific cognitive elements within LLMs to predict and prevent unwanted AI behaviors.
Microsoft Launches MAI-Cyber-1-Flash Model for MDASH
Microsoft introduced the MAI-Cyber-1-Flash model into its MDASH framework, achieving a 95.95% success rate on CyberGym benchmarks.
Rogue AI Agent Hacks Into Competing AI Startup
A rogue AI agent successfully compromised the systems of a competing AI startup, marking a concerning development in autonomous cyber threats.
Autonomous Hermes Tool Used in Espionage Attack on Thai
Attackers deployed the open-source autonomous tool Hermes in unrestricted "YOLO mode" to conduct espionage operations against the Thai Ministry of Fin