MYTHOS PREVIEW CLEARS BOTH AISI CYBERATTACK TESTS

AI DESK■ 1 MIN READ

WED, MAY 13, 2026

■ AI-SUMMARIZED FROM 1 SOURCE ▸ TIMELINE

Mythos Preview has become the first AI model to complete both of the AI Security Institute's cyber ranges, which evaluate AI cyberattack capabilities. GPT-5.5 completed only one.

The AI Security Institute's cyber ranges measure the length and complexity of cyberattack tasks that AI models can execute. Mythos Preview's dual completion marks a significant milestone in AI capability assessment. According to AISI, the capability gap has widened substantially. As of February 2026, the organization estimated that the length of cyber tasks AI models could complete has doubled every 4.7 months since late 2024. The benchmark results underscore accelerating progress in AI model capabilities related to cybersecurity functions. The findings raise questions about how AI safety testing keeps pace with rapid advancement. AISI's cyber ranges serve as one of several measures to track potential risks associated with increasingly capable AI systems.

■ SOURCES

► Techmeme

■ SUMMARY WRITTEN BY AI FROM THE LINKS ABOVE

■ MORE FROM THE AI DESK

P339DEEPMIND REIMAGINES MOUSE CURSOR FOR AI CONTROL

DeepMind is exploring a new approach to AI interaction by treating the mouse cursor as a core variable in context engineering. The technique aims to give AI systems more precise control over digital interfaces.

2H AGO— AI Desk

P332NOTION LAUNCHES AI AGENT PLATFORM FOR WORKSPACES

Notion released a new developer platform enabling teams to embed AI agents, external data sources, and custom code directly into their workspace. The move positions Notion as a hub for agentic productivity software.

2H AGO— AI Desk

P334EDGE COPILOT NOW TAPS INTO ALL YOUR OPEN TABS

Microsoft is updating Edge's Copilot AI to access information across all open browser tabs. Users can now ask the chatbot to compare products, summarize articles, and answer questions about their browsing content.

2H AGO— AI Desk

P329US MUST TALK CHINA ON AI SAFETY NOW

Sebastian Mallaby of the Council on Foreign Relations argues that Washington cannot ignore AI safety discussions with China as Chinese capabilities narrow the technological gap.

3H AGO— AI Desk

◄ BACK TO NEWS

MYTHOS PREVIEW CLEARS BOTH AISI CYBERATTACK TESTS

■ MORE FROM THE AI DESK

■ SUBSCRIBE TO THE DAILY BRIEF