EU AI Act — Article 13 Disclosure
AI Transparency Disclosure
WP-Claw uses AI agents to manage WordPress sites autonomously. This page documents our AI systems, their capabilities, limitations, and the human oversight mechanisms in place — as required by the EU AI Act.
Risk Classification
WP-Claw is classified as a limited-risk AI system under the EU AI Act. Our agents manage website operations (SEO, security headers, content, analytics) but do not:
- — Make decisions affecting fundamental rights
- — Process biometric data
- — Perform social scoring or profiling
- — Operate critical infrastructure
WP-Claw's primary obligation under the AI Act is transparency — ensuring users know they are interacting with AI and understand how the system works.
AI Model Providers
| Provider | Purpose | Data Routing |
|---|---|---|
| Anthropic (Claude) | Content generation, SEO analysis, task planning | EU-routed API |
| Google (Gemini) | Analytics interpretation, pattern detection | EU-routed API |
| OpenAI (GPT-4) | Code analysis, security scanning | EU-routed API |
All API calls are routed through EU-based infrastructure. No customer data is used for model training. Model providers process data under data processing agreements compliant with GDPR Article 28.
What the Agents Can Do
WP-Claw deploys up to 6 AI agents per WordPress site. Each agent has a defined scope:
Karim (Architect)
Plugin management, updates, site health, configuration
Lina (Scribe)
SEO metadata, content drafts, internal linking, schema markup
Bastien (Guardian)
Security headers, SSL monitoring, malware scanning, backups
Selma (Analyst)
Traffic analysis, Core Web Vitals, performance monitoring
Hugo (Commerce)
WooCommerce optimization, CRM, form management
Concierge
Live chat, visitor assistance, lead capture
Human Oversight Mechanisms
Tiered autonomy model
Every agent action is classified into 5 tiers: AUTO (read-only), AUTO+ (execute + notify), PROPOSE (execute with rollback window), CONFIRM (wait for human approval), BLOCK (never execute). High-risk actions always require human approval.
Real-time dashboard
Every action taken by every agent is visible in your dashboard in real time. You can see what each agent did, why, and the outcome — before and after.
Proposal system
Actions classified as PROPOSE or CONFIRM create a proposal visible in your dashboard. PROPOSE actions have a 15-minute rollback window. CONFIRM actions wait for your explicit approval.
Immutable audit log
Every API call, agent action, and system event is logged with timestamp, agent identity, action type, and outcome. Audit logs are retained for 90 days and available for export.
Known Limitations
- AI-generated content may contain factual errors. All content drafts should be reviewed before publishing.
- SEO recommendations are based on general best practices and may not account for niche-specific ranking factors.
- Security scanning detects known vulnerability patterns but cannot guarantee detection of zero-day exploits.
- The live chat agent (Concierge) cannot process payments, access customer accounts, or make commitments on your behalf.
- Agents operate within the WordPress environment and cannot modify server configuration, DNS settings, or hosting-level resources.
Questions about our AI systems? Contact us at [email protected]
Last updated: March 2026. This disclosure will be updated as our AI systems evolve.