Security Policy

Constrained autonomy

LORE-0 separates code generation from permission to act. Generated capabilities do not automatically receive unrestricted authority.

• No unrestricted payment, messaging, shell or external-write permissions by default.
• Public-web research is treated as a read-only network capability.
• Agent-on-Demand designs are screened for unsafe imports and side effects.
• Sensitive external effects are routed to composition or human review.
• Commercial execution is subject to authentication, rate limits and Profit Guard checks.

Acceptable-use Safety Gate

LORE-0 applies a deterministic Safety Gate before automated quote, payment, build, catalog publication and execution. Restricted categories include sexual services/explicit sexual content, child sexual exploitation, weapons/explosives, illicit drugs, alcohol/tobacco/nicotine commerce or promotion, fraud/cybercrime/theft, trafficking/exploitation, terrorism/violent-extremist support and other explicitly illegal activity.

Clearly facilitative requests are blocked. Sensitive requests that appear protective, compliance-oriented, academic, investigative or safety-related are not automatically executed or monetized; they require human review.

Read the LORE-0 Acceptable Use Policy →

Vulnerability reporting

Send suspected vulnerabilities to contacto@stardesign.cl with SECURITY in the subject.