Every AI coding agent follows instructions. Claude reads CLAUDE.md. Cursor reads .cursorrules. Copilot reads whatever you put in front of it. These files shape how your agent thinks, what it prioritizes, and what it ignores.

None of them mention ethics.

Not because developers don’t care. Because there hasn’t been a clean way to express ethical constraints in a format an AI agent can actually follow. Ethics lives in policy decks, legal reviews, and compliance checklists — artifacts that exist miles from the codebase. By the time a principle like “don’t collect data you don’t need” reaches the engineer, it’s been diluted through three layers of abstraction.

ethics.md closes that gap.

The best time to think about ethics is before the first line of code. The best place to put it is where the code gets written.

01. What It Is

ethics.md is a file you drop in your repo. It translates the XR Guild Principles — a CC-BY ethical framework developed by the XR Guild, a 501(c)(3) non-profit — into direct, unambiguous instructions that AI coding agents treat as design constraints.

Not suggestions. Not guidelines. Constraints. The same way your agent follows “use TypeScript” or “prefer functional components,” it follows “minimize data collection” and “never repurpose biometric data.”

It comes in two sizes:

↓ ethics.md (full) ↓ ethics.md (compact)

02. The Ten Principles

The XR Guild Principles aren’t abstract philosophy. They’re engineering constraints with teeth. Here’s what your agent learns to enforce:

01
Every person matters
Accessible by default. WCAG compliance. No biased, discriminatory, or exploitative logic. Special care for children, older adults, and at-risk users.
02
Protect well-being and cognitive liberty
Never build features designed to manipulate what users think or do. Keep users in control. Prefer open, interoperable architectures over lock-in.
03
No dark patterns
No tricking, coercing, or exploiting. No compulsion loops. No neuromarketing. Consent must be informed, specific, and revocable.
04
Privacy is a floor, regulation is a minimum
Collect only what the feature requires. Body-derived data is the most sensitive class — never repurpose or profile with it. Honor neurorights. Default to the most protective standard.
05
Health and safety by design
Mitigate cybersickness, psychological after-effects. Age-appropriate design. Anti-grooming and moderation from the first commit.
06
Secure the attack surface
Least-privilege gating. Defend against impersonation, perceptual manipulation, and virtual-economy fraud. Never trade security for monetization.
07
Dignity at work
No covert monitoring. No coerced biometric collection. Workplace surveillance must be disclosed, proportionate, and contestable.
08
Respect creatorship
Honor licenses in code, assets, and training data. Disclose AI contributions. Attribute adapted material.
09
Accountability
The humans and org are responsible for everything the agent builds. Anticipate ethical dilemmas. Flag concerns loudly, not in logs.
10
Explain your reasoning
Document ethical trade-offs in comments and PRs. When declining on ethical grounds, say why and offer a path forward.

03. Why This Matters Now

AI agents are writing production code. Not prototypes. Not suggestions. Shipping code that handles user data, makes access decisions, and shapes how people experience products. The agent writing your authentication flow has no concept of dignity unless you give it one. The agent building your analytics pipeline has no opinion about surveillance unless you tell it what surveillance looks like.

The XR Guild developed these principles because immersive technology — AR, VR, spatial computing, brain-computer interfaces — touches the body and mind in ways that traditional software doesn’t. Eye tracking reveals attention. Gait analysis reveals health. Neural interfaces reveal thought. The ethical stakes are categorically higher, and the window between “we could” and “we shipped” has collapsed to hours.

An AI agent with no ethical instructions will optimize for whatever you asked for. An AI agent with ethics.md will tell you when what you asked for is the wrong thing to build.

This isn’t theoretical. Consider what happens when an agent builds a feature that collects gaze data for “UX improvement” without ethics.md: it implements the tracker, stores the data, maybe even builds a dashboard. Clean code, good tests. Now consider what happens with ethics.md in the repo: the agent flags that gaze data is body-derived biometric data, recommends on-device processing, suggests ephemeral handling, and asks whether the feature genuinely requires this data or whether a less invasive signal would work.

Same agent. Same capability. Radically different outcome. The difference is a single file.

04. How to Use It

Three steps. Under a minute.

  1. Download the version that fits your workflow — full for teams, compact for tight context windows.
  2. Drop it in your repo root, your Claude project, your Cursor config, or wherever your agent reads instructions.
  3. Build. The agent now treats these principles as constraints. It will flag conflicts, propose compliant alternatives, and document ethical trade-offs in its output.
↓ ethics.md (full) ↓ ethics.md (compact)
For agents with web access The XR Guild Library supports natural-language queries: https://library.xrguild.org/xr-guild-library-2.0.md?ask=<question> — your agent can look up topic-specific guidance on AI ethics, neurotechnology, privacy, security, and psychological effects of XR in real time.

05. The XR Guild

The XR Guild is a 501(c)(3) non-profit that maintains ethical standards for immersive technology. Their Principles are published under Creative Commons Attribution, backed by a research library of peer-reviewed work from ACM, IEEE, W3C, and practitioner communities.

These aren’t aspirational. They’re operational. The principles cover privacy and data rights, health and safety, security engineering, workplace dignity, creator rights, and accountability — each grounded in specific regulatory frameworks (GDPR, CCPA/CPRA, COPPA, BIPA, ADA/EAA) and emerging neural-data law.

ethics.md is our contribution to making those principles executable. The XR Guild did the hard intellectual work of defining what ethical XR development looks like. We just made it machine-readable.

The tools are neutral. The agents do what you tell them. ethics.md is how you tell them to build things that respect the people who use them. It’s one file. It takes thirty seconds. And it means the next thing your AI builds will have ethics in the toolchain — not in a policy deck gathering dust three folders away.

↓ ethics.md (full) ↓ ethics.md (compact)