AISI Evaluation Shows Claude Mythos Preview's Cyber Capabilities in CTF and Multi-Step Attacks
The AI Security Institute (AISI) conducted cyber evaluations of Anthropic's Claude Mythos Preview, assessing its performance on capture-the-flag challenges and multi-step attack simulations. The model showed significant improvement over previous frontier models in cybersecurity capabilities.
Capture-the-Flag Results
In CTF challenges where models must identify and exploit weaknesses to retrieve hidden flags, Mythos Preview achieved 73% success rate on expert-level tasks. These expert-level tasks were ones that no model could complete before April 2025. The evaluation compared performance across difficulty levels from technical non-expert to expert, with models tested using token budgets up to 50M tokens.
Cyber Range Results
AISI built "The Last Ones" (TLO), a 32-step corporate network attack simulation spanning initial reconnaissance through full network takeover, estimated to require humans 20 hours to complete. Claude Mythos Preview was the first model to solve TLO from start to finish, succeeding in 3 out of 10 attempts. Across all attempts, the model completed an average of 22 out of 32 steps.
Claude Opus 4.6 was the next best performing model, completing an average of 16 steps. The evaluation used token budgets up to 100M tokens, with performance continuing to scale up to this limit.
Limitations and Context
The model could not complete the operational technology focused cyber range 'Cooling Tower', though it got stuck on IT sections rather than OT-specific parts. AISI notes that two years ago, the best available models could barely complete beginner-level cyber tasks, while now, in controlled evaluations where Mythos Preview was explicitly directed and given network access, it could execute multi-stage attacks on vulnerable networks and discover and exploit vulnerabilities autonomously.
📖 Read the full source: HN AI Agents
👀 See Also
Claude Android App Reportedly Reads Clipboard Without Explicit User Action
A user reports that the Claude Android app analyzed code from their clipboard without them pasting it, with Claude identifying the file as pasted_text_b4a56202-3d12-43c8-aa31-a39367a9a354.txt. The behavior couldn't be reproduced in subsequent tests.
ClawSecure: Security Platform for OpenClaw Ecosystem with 3-Layer Audit and Real-Time Monitoring
ClawSecure is a dedicated security platform for OpenClaw that performs 3-layer security audits, real-time monitoring with SHA-256 hash tracking every 12 hours, and provides full OWASP ASI coverage. It has audited 3,000+ popular skills and is free to use with no signup required.
Trojan found in Claude Flow repository skill.md files
A GitHub repository containing Claude Flow skill files was found to contain a Trojan identified as JS/CrypoStealz.AE!MTB. The malware triggered automatically when an AI-based IDE opened the folder to read the markdown files.
Zero-Trust OpenClaw Architecture Adds Pre-Execution Authorization and Post-Execution Verification
An open-source architecture for OpenClaw adds two security checkpoints: a Rust sidecar that intercepts tool calls before execution with sub-millisecond authorization overhead, and deterministic post-execution verification using assertions instead of LLM judgment. The system includes tracing with DOM snapshots and screenshots, plus a DOM compression skill that reduces token usage by 90-99%.