← Back to Governance

Easylab AI -- Regulatory Documentation

EASYCLAW SYSTEM CARD

Technical Documentation -- Article 11, Regulation (EU) 2024/1689

Version 1.0|March 2026|EC-SC-2026-001

1.GENERAL DESCRIPTION

Annex IV(1)

--System name: EasyClaw
--Provider: Easylab AI
--Classification: General-purpose autonomous agent platform -- AI deployer application (Article 26). Risk classification is determined per deployment context, not globally.
--Underlying model: OpenAI GPT-5.4 (GPAI model, provided by OpenAI)
--Framework: OpenClaw orchestration layer
--Intended purpose: Autonomous AI agent platform for B2B deployment

2.SYSTEM ARCHITECTURE

Annex IV(2)

--Core: OpenAI GPT-5.4 LLM
--Orchestration: OpenClaw framework (Node.js)
--Memory: Qdrant vector database, Mem0, Cognee, SQLite
--Communication: Multi-channel (Telegram, iMessage, Email)
--Security: exec-guardian daemon, chflags file protection, PIN authentication
--Monitoring: Real-time dashboard (8 panels), Bridge Governor, Council Governor

3.CAPABILITIES

Annex IV(3)

--Autonomous task execution and scheduling
--Persistent memory with extraction and distillation cycles
--Multi-channel bidirectional messaging
--System command execution within security boundaries
--File management and code generation
--Web research and data analysis
--Inter-agent communication (Bridge protocol)

4.LIMITATIONS

Annex IV(4)

--No real-time autonomous internet browsing without configured tools
--No biometric data processing capability
--No image, audio, or video generation
--Bounded by exec-guardian operational constraints
--Requires human PIN approval for sensitive system operations
--Cannot override chflags-protected configuration files
--Inference quality dependent on underlying LLM capabilities
--No guaranteed factual accuracy -- outputs should be verified

5.RISK MANAGEMENT

Art. 9

--14-layer security architecture
--Rate limiting on inter-agent communications (50 msg/direction)
--Automated memory distillation to prevent unbounded data growth
--Real-time security monitoring and alerting
--Human-in-the-loop for critical operations (exec-guardian PIN system)

6.DATA PROCESSING

Art. 10

--All data stored locally on deployment machine
--LLM API calls to OpenAI for inference (subject to OpenAI's data policies)
--No training on deployer data by the provider
--Configurable data retention policies
--GDPR-compliant data deletion capabilities

7.PERFORMANCE METRICS

Annex IV(5)

--System uptime: 99.7% (monitored)
--Trust score: Real-time calculation based on operational metrics
--Security incident rate: Logged and tracked via dashboard
--Memory accuracy: Validated through distillation process

8.HUMAN OVERSIGHT MEASURES

Art. 14

--Exec Guardian daemon with PIN authentication
--Real-time monitoring dashboard
--Kill switch via Telegram or dashboard
--Bridge Governor rate limiting
--Council Governor multi-agent consensus
--Comprehensive audit logging

9.CONFORMITY

Art. 43

--Self-assessment conducted per Article 43(2)
--Quality management system per Article 17
--Post-market monitoring per Article 72
--This documentation maintained per Article 18

System Card - Technical Documentation Article 11 | EasyClaw by Easylab AI