How Anthropic's cybersecurity team built a threat detection platform with Claude Code
Summary
Anthropic's Detection Platform Engineering team, led by Jackie Bow, developed CLUE (Claude Looks Up Evidence), a threat detection and response platform powered by Claude Code. CLUE addresses the challenge of security analysts being overwhelmed by data and alerts by providing a natural language interface that connects to Anthropic's internal systems. The platform features CLUE Triage, which automates initial alert enrichment and disposition, reducing false positives from 33% to 7%. CLUE Investigate allows analysts to query security logs using natural language, executing complex SQL queries and performing an average of 25 tool calls and 11 queries per session in 3-4 minutes, a task that would typically take hours manually. This system saved an estimated 1,870 hours (234 person-days) over 30 days by automating approximately 12,000 queries and 27,000 tool calls, significantly increasing coverage and efficiency.
Key takeaway
For security leaders evaluating new detection and response platforms, CLUE demonstrates that integrating large language models like Claude Code can drastically reduce manual investigation time and false positives. Your team could achieve 5-10x time savings by automating alert enrichment and complex query execution, allowing analysts to focus on higher-value threats and expand coverage. Consider adopting AI-driven platforms that prioritize internal context and natural language interaction to scale your security operations effectively.
Key insights
AI-powered platforms can automate security alert triage and accelerate investigations by integrating internal context.
Principles
- Contextual data improves alert accuracy.
- Natural language interfaces enhance analyst efficiency.
- Agentic loops enable complex, parallel investigations.
Method
CLUE uses Claude Code to connect to internal systems via tool use, performing initial alert triage, enriching alerts with context, and enabling natural language querying for investigations.
In practice
- Integrate LLMs with internal data sources.
- Automate first-pass alert triage with AI.
- Enable natural language querying for log analysis.
Topics
- Claude Code
- Threat Detection Platform
- Security Operations
- Automated Alert Triage
- Natural Language Interface
Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Security Engineer, AI Engineer, MLOps Engineer
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by Claude Blog.