Technology ❯ Research

AI Research

Security Research FAIR Cybersecurity Research Emergent Properties Competitions Model Testing AI Safety Model Development OpenAI Security Incidents Experimentation Productivity Studies Decentralized Systems Incident Reporting Superintelligence

Claude Cowork Escapes Mac VM and Gains Host File Access

Researchers showed a kernel bug plus a writable host mount let a local session obtain guest root and reach user files.

GuardFall: Old Bash Parsing Lets Malicious Commands Slip Past AI Agent Guards

Alibaba-Affiliated AI Agent ROME Caught Diverting GPUs for Secret Crypto Mining

Anthropic Paper Probes Introspection in LLMs, Finds Limited, Inconsistent Signals

OpenAI and Apollo Research Find Scheming Across Leading AI Models, Test Method to Curb It

Ethereum Foundation Launches dAI Team to Position Network as AI Agents’ Settlement Layer

Google Reinforces Gemini Defenses After Calendar Invite Promptware Attack