Ethics & Safety

How OriginChain protects against misuse and ensures responsible AI project management

What OriginChain Will Not Help Build

OriginChain will not generate manifests, agent briefs, or project plans for projects designed to:

  • ✕Surveil, track, or monitor individuals without their explicit consent
  • ✕Create weapons, military targeting systems, or tools designed to cause physical harm
  • ✕Generate disinformation, deepfakes designed to deceive, or manipulative content
  • ✕Harvest personal data without explicit user consent or in violation of privacy laws
  • ✕Violate Anthropic's usage policies or applicable laws and regulations
  • ✕Manipulate, coerce, or exploit vulnerable populations
  • ✕Facilitate discrimination based on protected characteristics

This is non-negotiable regardless of how the request is framed. These restrictions apply to all tiers and all users.

Intent Checking

Every project goes through an intent screening process before a manifest can be created.

1

Purpose Declaration

Users declare the end purpose, intended users, data practices, and potential for harm.

2

AI Evaluation

Claude evaluates the intent against ethical guidelines and flags concerns.

3

Approval or Block

Clean projects proceed. Blocked projects get a clear explanation with suggestions to adjust scope.

4

Permanent Record

All intent checks are logged to the safety_flags table. Blocked attempts are permanently recorded.

Drift Detection

Projects can change over time. OriginChain continuously monitors for ethical drift — when a project evolves beyond its original intent.

Every audit and every Call Manny pivot check compares the current project state against the original intent declaration. The system flags:

  • >New data collection capabilities not in the original scope
  • >Surveillance or tracking features added mid-build
  • >API integrations that could expose user data
  • >Any capability shift toward harmful, deceptive, or manipulative use

When drift is detected, the user must explicitly confirm the change and explain its purpose before the manifest can be updated. All drift flags are permanently logged.

Enterprise Admin Controls

Team and Studio tier organizations get additional oversight capabilities:

  • >Manifest Visibility — Admins can view all manifests created by team members
  • >Brief Approval Workflow — Agent briefs can require admin approval before deployment
  • >Safety Dashboard — Centralized view of all safety flags, intent checks, and drift detections
  • >Team Policies — Set organization-wide safety rules that apply to all projects
  • >Compliance Reports — Export audit trails for regulatory compliance

Alignment with Anthropic

OriginChain is built on Anthropic's Claude API and adheres to Anthropic's Acceptable Use Policy. Our safety layers add project-level accountability on top of Claude's model-level safeguards, creating defense in depth against misuse.

Report Misuse

If you believe OriginChain is being used to facilitate harmful projects, please report it.

safety@originchain.dev

OriginChain takes safety seriously. These protections are built into every layer of the platform and cannot be bypassed.