Live Nov 18 · Cohort 3Nov 18, 19:00 MSK — free hands-on AI agents live workshop · Live Cohort 3 starts Nov 23Join Nov 18 Live →Cohort 3 (Nov 23)
ReleaseInsight · On RadarHabr AI→ original

AI helped write the code: the studio needs to separate rights to 3 parts of the product

The client wants all rights to a product the team built with AI. But it includes an MIT-licensed library, the studio’s own platform and a new adapter: a promise that “everything is ours, with no third-party materials” already conflicts with what is being delivered. Before signing, separate these components and describe what the client will receive. A subscription to an assistant does not, on its own, establish rights to the entire product.

AI-processed from Habr AI; edited by Hamidun News
AI helped write the code: the studio needs to separate rights to 3 parts of the product
Source: Habr AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceHabr AI ↗
Editor-in-Chief
Key metric #1Document the release, dependencies and licenses included in the deliverables.
Key metric #2Agree separately on rights to the platform and its general-purpose improvements.
Key metric #3Check whether you have permission to share the client’s code and data with the assistant.

The client has sent over a contract: assign all rights to the work and confirm that it contains no third-party materials. The team wrote the integration using Cursor, Claude Code or Codex and is now trying to determine whether it can sign. Start with the contents of the release. If it includes someone else’s library, the studio’s own platform and a new adapter, a single phrase about “all rights” conceals three separate agreements.

For a developer, this is a component-tracking task. For the studio’s founder, the question is whether the team can reuse its work on the next project. For the client, it is whether they can maintain the system they paid for after parting ways with the contractor. AI involvement raises additional questions, but much of the work is familiar from conventional software development.

First, break down the deliverables

Consider a studio connecting a client’s payment system to its own platform. The project introduces an adapter, an MIT-licensed library and a faster payment-processing queue. The assistant helped write the adapter and rework the queue.

The library remains a third-party component. Permission to use it in a commercial product does not make the studio the holder of exclusive rights to it. The statement “there are no third-party materials” therefore needs correcting regardless of who chose the package—a person or an assistant.

The platform is more complicated. A clause stating that “pre-existing work remains with the contractor” leaves the improved queue unresolved: it was created during the project. If the studio plans to reuse that improvement, the parties need to agree separately on how it will be treated.

That leaves three subjects for discussion: third-party components, the platform and its general-purpose improvements, and client-specific development. The boundary may run through a single file. Organizing code into folders is not enough: the contract must describe the agreed allocation of rights.

What to check in the code and the assistant’s terms

Start with a specific release: source code, build, documentation and dependencies. That way, both parties will be discussing the same deliverables rather than a working branch that changes every day.

The source material proposes two tools: Syft for a software bill of materials, or SBOM, and ScanCode Toolkit for finding license and authorship information in files. This inventory helps build a picture of the deliverables. A third-party snippet embedded in your own file will require additional review of the code and change history. An empty scanner report does not resolve every question about the material’s provenance.

The next document to examine is the terms of the AI service used. The article highlights a fundamental distinction: provisions on user rights govern the relationship between the user and the assistant’s provider. They do not automatically establish rights to libraries, client materials or the complete assembled product.

Any protection against claims, along with its exclusions, requires separate scrutiny. That means identifying the exact product, plan and applicable version of the terms. A note saying “we have a paid subscription” is not enough to compare obligations: the studio’s promises to its client may be broader than the AI provider’s commitments.

Give the client clear permissions

Keeping the platform with the studio is only half the agreement. The client needs permissions sufficient for its use case: running the system, fixing bugs, hiring another team and giving that team access to the source code. Selling access to outside companies also requires a separate discussion.

Before connecting the assistant, check whether you may share the client’s source code, specifications and data with it. Disabling training does not, by itself, answer whether that transfer is permitted. A workable arrangement might include an agreed tool, restricted file access and synthetic records in place of real ones.

The practical outcome of this review is an appendix listing the deliverables and the agreed rights to their components. It needs to be checked against the contract, arrangements with developers and applicable law. If the release contains a third-party library while the contract still promises a complete absence of third-party materials, it is too early to sign that wording.

Story

Key Questions Answered

What are the key details on First, break down the deliverables?

Consider a studio connecting a client’s payment system to its own platform. The project introduces an adapter, an MIT-licensed library and a faster payment-processing queue. The assistant helped write the adapter and rework the queue. The library remains a third-party component. Permission to use it in a commercial product does not make the studio the holder of exclusive rights to it.

What to check in the code and the assistant’s terms?

Start with a specific release: source code, build, documentation and dependencies. That way, both parties will be discussing the same deliverables rather than a working branch that changes every day. The source material proposes two tools: Syft for a software bill of materials, or SBOM, and ScanCode Toolkit for finding license and authorship information in files. This inventory helps build a picture of the deliverables.

What are the key details on Give the client clear permissions?

Keeping the platform with the studio is only half the agreement. The client needs permissions sufficient for its use case: running the system, fixing bugs, hiring another team and giving that team access to the source code. Selling access to outside companies also requires a separate discussion. Before connecting the assistant, check whether you may share the client’s source code, specifications and data with it. Disabling training does not, by itself, answer whether that transfer is permitted.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…