Who owns what a local AI model writes? The model licences we checked do not claim your output. Indian law has no settled answer yet.
Short answer. Probably the office that used the model, but Indian law is unsettled. The Copyright Act treats the person who causes a computer-generated work to be created as its author (section 2(d)(vi)), and the employer is first owner of work made in the course of employment unless a contract says otherwise (section 17(c)). Whether text produced with little human input is protected at all is not settled: DPIIT’s promised second paper on AI outputs had not appeared by October 2026, and the US Copyright Office says prompts alone do not give enough human control for protection. The licences of the models we checked do not claim your outputs. Put ownership, review and confidentiality in the engagement letter or client contract. This is general information, not legal advice.
01
What Indian law says today
The Copyright Act, 1957 defines the author of a computer-generated literary, dramatic, musical or artistic work as “the person who causes the work to be created” (section 2(d)(vi)). The author is the first owner (section 17), but a work made in the course of employment under a contract of service belongs first to the employer, in the absence of an agreement to the contrary (section 17(c)).
The Act does not say whether text a model generates from a short prompt is a protectable work, or who “causes” it. DPIIT’s committee said in its December 2025 working paper that a second part would cover the copyright status of AI-generated outputs, including who should be the author. On 5 October 2026 Hindustan Times reported that this second paper was still not ready, nearly ten months after the first, and that the committee had not met since June.
02
What the US Copyright Office concluded
US law is not Indian law, but the Copyright Office’s January 2025 report is the most detailed official view of where authorship lines may fall.
Copyright does not extend to purely AI-generated material, or to material where there is insufficient human control over the expressive elements.
On current generally available technology, prompts alone do not provide sufficient control.
Human authors keep copyright in their own expression that is perceptible in an AI output, and in the creative selection, coordination or arrangement of material, or creative modifications of it.
Whether human contributions are enough is decided case by case.
03
What the model licences say
Licence or terms
On outputs
Other conditions that touch outputs
Apache 2.0 (Qwen3.8, Qwen3, Gemma 4, gpt-oss)
Contains no clause giving the licensor rights in what you generate
The work is provided “as is”, so responsibility for outputs is yours
Gemma Terms of Use (Gemma 3 and earlier)
“Google claims no rights in Outputs you generate using Gemma”; you and your users are solely responsible for outputs
A model trained on Gemma outputs counts as a Model Derivative under the terms
Llama 3.3 Community License
No clause claiming ownership of outputs; the warranty disclaimer covers any output and results
If you use outputs to improve an AI model you distribute, its name must begin with “Llama”
The customer retains ownership of input and owns all output; OpenAI assigns any rights it has
OpenAI uses customer content only to provide the services, comply with law, enforce its policies and prevent abuse
Microsoft Copilot
Microsoft does not claim ownership of output, and makes no determination on whether output is copyright protected or enforceable against others
Microsoft defends commercial customers against copyright claims if they used the built-in guardrails
04
What to put in writing
In engagement letters and contracts, say whether AI tools may be used, which ones, and who owns the final work product.
Keep human authorship visible: record who drafted, who edited and who approved. More human selection and revision gives a stronger claim.
Check the model licence for conditions that follow the output, such as naming rules or a prohibited-use policy.
Do not feed client confidences to a tool whose terms let the vendor use them.
For anything you must be able to protect, such as a brand or a publication, rely on human-authored expression and keep the drafts.
05
Where ANVE fits
ANVE logs every model call with the user, app, model and whether it ran locally or in the cloud, and every confirmed meeting action records who confirmed its owner. That helps an office show who prompted, who edited and who approved. It is a record, not a legal opinion, and ANVE is pre-release.
06
Questions people ask
Can an AI system be the author of a work in India?
The Copyright Act names a person as the author of a computer-generated work: “the person who causes the work to be created”. The DPIIT paper that is meant to address authorship of AI outputs is still pending as of October 2026, so ask counsel before relying on a view.
Does the model vendor own my outputs?
Not under the terms checked here. Google’s Gemma terms say it claims no rights in outputs, Apache 2.0 has no such clause, and OpenAI’s services agreement says the customer owns all output.
Can we claim copyright in a document drafted by AI and edited by staff?
The US Copyright Office says human-authored expression and creative selection or arrangement can be protected. Indian law has no settled position. Keep drafts and a record of the human work.
Does running the model locally change ownership?
Not much: the licence and the law are the same. Running locally matters for confidentiality, because prompts and drafts stay on your device.
07
What is not real yet
No sealed ANVE unit exists; every image of the appliance on this site is a labelled render. The software runs today on a development machine. No customers yet, no certifications, no delivery date, and the final price is not announced.
Reservations are ₹999 in India or US$12 elsewhere, refundable in full until you confirm an order in writing (terms). Offices that cannot use cloud AI can apply to be one of ten design partners.
