AI and Copyright in Nepal: Who Owns Computer-Generated Art and Content?

The Act's human-author premise

The Copyright Act 2059 protects original works and vests them in the author — and its architecture (moral rights of attribution and integrity, life-plus-50 terms, authorship of joint works) is built entirely around natural persons. There is no provision deeming AI output a work, no provision vesting machine-generated content in the tool's operator, nothing analogous to the UK's computer-generated-works provision. Nepal is also not a party to the WIPO Internet Treaties, and its AI-copyright posture is, so far, whatever the general doctrine yields.

The doctrinal result: protection analysis for AI output reduces to the human contribution. Where a person's creative choices — the prompt design, the iterative selection, the edits, the arrangement into a final work — constitute original expression, the human is the author of that expression. Where output is machine-generated with no meaningful human creative input, there is likely no Nepalese copyright in it at all — which means anyone may copy it, including your competitors.

• Human-author architecture throughout — no AI provision exists
• Human creative direction → protectable as the human's work
• Pure machine output → likely no copyright; free for anyone

The three scenarios, worked

Scenario one — AI-assisted with real direction. A designer prompts, iterates, selects, retouches and composes a campaign visual. The human contribution is substantial and traceable; the work is protectable as the designer's — and the studio owns it only with a written assignment (the creator-first default applies as always). Scenario two — AI-generated with minimal input. 'A logo for a tea company' → output used as-is. The human input is a request, not expression; protection is doubtful, and the brand consequence is real: a logo you may not own can become the face of your business — and you cannot cleanly register or enforce a device mark whose copyright you don't hold. Scenario three — AI as a tool inside human work. Code co-written with an assistant inside a developer's architecture, copy drafted and substantially edited by a writer. The human creative process dominates; the work is protectable; document the process.

The through-line: the difference between one and three is documentation. Studios that record prompts, iterations and edits own the argument; studios that don't own a dispute.

The contracting layer: four documents that end the argument

1. AI-tool terms. The provider's terms decide what you may do with outputs (ownership grants, commercial-use rights) — read them per tool, and prefer providers granting output rights for commercial use. 2. Employment/contractor assignments — updated to cover AI-assisted works expressly: 'all works created… including works created with AI-assisted tools.' 3. Client agreements — disclose AI use where material, warrant the human creative contribution and non-infringement, and assign deliverables in present terms. 4. Indemnity awareness — AI tools trained on third-party works create infringement risk that lands somewhere; know whether your contract allocates it to the provider (rarely), your client (by silence), or you.

For the registration question: voluntary registration at the Copyright Registrar's Office remains available for protectable works — and the application's authorship statements should reflect the human author honestly. Overclaiming machine output as solely human-authored work is both inaccurate and, in a dispute, fatal.

Where this is heading

Internationally, the AI-copyright question is unsettled and moving: jurisdictions are splitting between human-contribution doctrines, new computer-generated-works categories, and outright denial of protection — and Nepal's framework will eventually face the same choices, whether through amendment, the Bill-era reform cycle, or case law forced by a dispute. Nepal's likely path, given its human-author architecture, is the human-contribution doctrine: protect the person's expression, not the machine's output.

For Nepalese creators and studios, the current advice is stable regardless of where the doctrine lands: direct the work, document the direction, and paper the ownership. Creators who do all three own their AI-assisted output under any plausible future rule — and can prove it under today's.