AI-Generated Content & Code Copyright Law In Pakistan: Authorship, Ownership & Infringement

Who owns code and creative works generated by ChatGPT, GitHub Copilot, or Midjourney under Pakistani law? Deep analysis of the human authorship requirement under the Copyright Ordinance 1962.

By Syed Asad Hussain Zaidi · Advocate High Court & Tax Consultant · 9 September 2026

Human Authorship Prerequisite: Under Section 2(d) of the Copyright Ordinance 1962, an "author" is strictly defined as a natural person. Pure artificial intelligence algorithms and autonomous AI agents have no legal personality and cannot hold copyright in Pakistan. The Public Domain Default: Works created entirely by an automated AI model without substantial creative intervention by a human being fall immediately into the public domain and cannot be exclusively protected against third-party copying. The Human-AI Collaborative Standard: Where a human software engineer or creative professional uses AI as an assistive tool—exercising creative selection, prompt engineering, architectural editing, and substantial modification—copyright subsists in the human author's original creative contributions. Corporate Risk Exposure: Incorporating AI-generated code snippets into commercial software without reviewing the underlying foundation model terms of service exposes tech enterprises to copyright infringement claims and warranty breaches in B2B vendor contracts. The AI Revolution Meets Statutory Copyright Jurisprudence The widespread adoption of generative artificial intelligence models—including OpenAI's ChatGPT, Anthropic's Claude, GitHub Copilot, Cursor, Midjourney, and Stable Diffusion—has revolutionized software development, digital publishing, and commercial content generation across Pakistan's tech ecosystem. Software developers now generate complex backend functions using AI prompts, digital agencies produce commercial illustrations in seconds, and content writers deploy automated workflows to produce high-volume marketing copy. However, this explosive operational efficiency has created profound legal ambiguity: Can an AI-generated software application, digital illustration, or commercial article receive legal copyright protection under Pakistani law? And who faces liability when an AI model outputs code or text copied from third-party protected repositories? Navigating these challenges requires examining the fundamental principles of originality, authorship, and intellectual property ownership embedded within the Copyright Ordinance, 1962. --- The Core Statutory Barrier: Section 2(d) and the Natural Person Rule Pakistani copyright law is rooted in human creative intellectual labor. Section 2(d) of the Copyright Ordinance 1962 defines an "author" across various categories: In relation to a literary or dramatic work: the author of the work; In relation to a musical work: the composer; In relation to an artistic work other than a photograph: the artist; In relation to a photograph: the person taking the photograph. In legal interpretation, an "author", "composer", or "artist" has consistently been held by the High Courts of Pakistan to denote a natural human person possessing legal capacity and intellectual intent. An artificial intelligence system—regardless of how sophisticated its neural network parameters may be—possesses no legal personhood, cannot execute contracts, cannot be sued for damages, and cannot own property. The Consequence of Zero Human Authorship: If a piece of software or digital artwork is generated 100% autonomously by an AI model in response to a single generic prompt (e.g., "Write an inventory management software in Python"), no copyright is created under Pakistani law. The resulting output enters the public domain immediately upon generation, meaning any competitor can freely copy, modify, and resell the code without infringing any intellectual property rights. --- The Human-Assisted Standard: When Does Copyright Attach? To secure legal copyright protection for software or content developed using AI tools, the human creator must demonstrate substantial human intellectual contribution. Pakistani jurisprudence follows the established common-law doctrine of "skill, judgment, and labor": | Level of AI Involvement | Legal Description | Copyright Status in Pakistan | | :--- | :--- | :--- | | Pure Autonomous Generation | User provides simple prompt; AI outputs entire function | No Copyright: Falls into public domain | | Iterative Human-AI Collaboration | Human designs system architecture, refines prompts, debugs logic | Protected: Copyright vests in the human developer | | AI as Assistive Copilot | Human writes core code; AI assists with autocompletion and syntax | Fully Protected: Standard human author copyright | | AI Infringing Output | AI replicates protected code snippets from third-party training data | Infringement Risk: Human user/firm strictly liable | Securing Protection for AI-Assisted Software: Document the Creative Architecture: Maintain clear version control (Git commit history) demonstrating how human software architects conceived the database schema, API routing, custom business logic, and security rules. Substantial Code Alteration: Ensure that AI-suggested code snippets are reviewed, customized, integrated, and enhanced by human developers rather than compiled wholesale. Document Prompt Workflows: Maintain records of specialized prompts and proprietary training fine-tuning that reflect human intellectual direction. --- Commercial Liability: The Third-Party Training Data Trap Generative AI models do not synthesize code or images from nothingness; they are trained on massive datasets scraped from the public web, including open-source repositories licensed under copyleft terms (GPL, AGPL) and proprietary copyrighted content. If an AI code assistant reproduces a 200-line block of proprietary code verbatim from an existing copyrighted software package, and your developers incorporate that snippet into a commercial enterprise SaaS product: Direct Infringement Liability: Under Section 56 of the Copyright Ordinance 1962, copyright infringement is an objective statutory offense. Ignorance is no defense; the fact that an AI generated the code does not shield your company from liability if the code is demonstrably an unauthorized copy of another party's protected work. Enterprise Warranty Breaches: Standard B2B software vendor contracts contain an explicit Intellectual Property Warranty certifying that the software provided is entirely original and free from third-party infringement. If an enterprise client gets sued by a copyright holder due to AI-plagiarized code inside your deliverable, your company will be liable for catastrophic indemnification claims. --- TaxCalc Advisory Insights: Mandatory Corporate AI Use Policies Every Pakistani tech company, digital marketing agency, and software house must immediately institute a formal Corporate Generative AI Policy. This policy should explicitly prohibit employees from pasting confidential client source code, customer personal data, or trade secrets into public AI tools where training data retention is enabled. Furthermore, developers must be contractually required to disclose AI usage and run automated code-origin scanning tools before merging pull requests into production branches. Institutional investors and foreign acquirers routinely require AI disclosure representations during M&A technical due diligence. Having documented AI governance safeguards your corporate valuation. --- Registering Works Created with AI Assistance at IPO-Pakistan When applying for copyright registration at the Copyright Office of IPO-Pakistan: State the Human Author Clearly: The applicant must always name a natural human person (or their corporate employer under a work-for-hire agreement) as the author. Never list an AI platform or algorithm as an author or co-author. Disclose Collaborative Nature in the Statement of Work: Clearly identify the human-authored components (system design, user interface, custom logic, integration algorithms) as the subject matter of protection. Execute Proper Internal Assignments: Ensure all software developers, content writers, and designers sign employment contracts explicitly assigning all IP created during employment—including works created using company-provided AI tools—to the company. --- Frequently Asked Questions (FAQs) Can an employer claim copyright over content created by an employee using ChatGPT? Yes. Under Section 13 of the Copyright Ordinance 1962, works created by an employee during the course of employment under a contract of service belong to the employer, provided the employee contributed sufficient human creative effort to satisfy the legal threshold of originality. Can an AI platform claim ownership of the code it generates for me? Under the Terms of Service of major commercial providers (such as OpenAI and Anthropic), the platform assigns all of its right, title, and interest in and to the output to the paying user. However, this contractual assignment only transfers whatever rights the platform possessed in the first place—if the output is uncopyrightable under local statutory law, the user receives an unprotectable work. How do intellectual property rights impact company valuation and tax planning? Proprietary intellectual property is an intangible capital asset. Developing and capitalizing original software allows companies to claim statutory amortization allowances under Section 24 of the Income Tax Ordinance 2001. Explore our <a href="/corporate-tax-calculator">Corporate Tax Calculator</a> to understand corporate asset deductions. --- The Contractual Defense: Drafting AI Representations and Warranties Because generative AI models introduce inherent risks of third-party copyright infringement, Pakistani software development firms serving international enterprise clients must master the art of drafting balanced AI Representations and Warranties in Master Service Agreements (MSAs): Balanced Client-Facing Clause: "The Developer represents that in the development of the Deliverables, it may utilize industry-standard automated coding assistance tools (including GitHub Copilot and equivalent commercial IDE plugins). The Developer covenants that all Deliverables have been reviewed, verified, customized, and architected by qualified human software engineers to ensure that the code is original and does not knowingly incorporate proprietary third-party code in violation of applicable licenses." What to Strictly Avoid in Client Agreements: Never Agree to Unconditional Absolute IP Indemnities: If a foreign client demands an unqualified guarantee that "the Deliverables will never infringe any intellectual property right anywhere in the world", and your engineers used AI copilots, refuse the clause. Insist on limiting the indemnity to "infringement known to the developer or resulting from gross negligence", and cap total liability to the contract value. --- Protection of AI Prompts and Fine-Tuning as Trade Secrets While raw text outputs from a public AI platform may be ineligible for statutory copyright registration, the underlying system prompts, fine-tuning datasets, agent architectures, and specialized RAG pipelines represent valuable proprietary intellectual property. Legal Protection under Common Law Trade Secrets: Confidentiality Agreements (NDAs): Treat complex prompt chains, evaluation benchmarks, and domain-specific dataset embeddings as protected confidential trade secrets under the Contract Act 1872. Restrictive Covenants for Prompt Engineers: Ensure technical staff execute enforceable Non-Disclosure and Non-Compete agreements prohibiting them from taking proprietary system prompts to competitor firms. Database Rights: Under Section 2(p) of the Copyright Ordinance 1962, curated databases of proprietary domain training data enjoy independent copyright protection as compilation literary works. --- Ethical and Regulatory Trajectory: What Tech Leaders Must Monitor The Ministry of Information Technology & Telecommunication (MoITT) and the Intellectual Property Organization of Pakistan (IPO-Pakistan) are actively drafting national artificial intelligence policy frameworks. Future statutory revisions are expected to address: Compulsory watermarking for commercial AI-generated imagery and video content; Explicit statutory recognition of computer-generated works along the lines of the UK Copyright, Designs and Patents Act 1988 (CDPA Section 9(3)), where the author is deemed to be the person who made the arrangements necessary for the creation of the work; Strict compliance mandates under the Prevention of Electronic Crimes Act 2016 (PECA) prohibiting deepfake generation, identity theft, and algorithmic consumer deception. --- The Contractual Defense: Drafting AI Representations and Warranties Because generative AI models introduce inherent risks of third-party copyright infringement, Pakistani software development firms serving international enterprise clients must master the art of drafting balanced AI Representations and Warranties in Master Service Agreements (MSAs): Balanced Client-Facing Clause: "The Developer represents that in the development of the Deliverables, it may utilize industry-standard automated coding assistance tools (including GitHub Copilot and equivalent commercial IDE plugins). The Developer covenants that all Deliverables have been reviewed, verified, customized, and architected by qualified human software engineers to ensure that the code is original and does not knowingly incorporate proprietary third-party code in violation of applicable licenses." What to Strictly Avoid in Client Agreements: Never Agree to Unconditional Absolute IP Indemnities: If a foreign client demands an unqualified guarantee that "the Deliverables will never infringe any intellectual property right anywhere in the world", and your engineers used AI copilots, refuse the clause. Insist on limiting the indemnity to "infringement known to the developer or resulting from gross negligence", and cap total liability to the contract value. --- Protection of AI Prompts and Fine-Tuning as Trade Secrets While raw text outputs from a public AI platform may be ineligible for statutory copyright registration, the underlying system prompts, fine-tuning datasets, agent architectures, and specialized RAG pipelines represent valuable proprietary intellectual property. Legal Protection under Common Law Trade Secrets: Confidentiality Agreements (NDAs): Treat complex prompt chains, evaluation benchmarks, and domain-specific dataset embeddings as protected confidential trade secrets under the Contract Act 1872. Restrictive Covenants for Prompt Engineers: Ensure technical staff execute enforceable Non-Disclosure and Non-Compete agreements prohibiting them from taking proprietary system prompts to competitor firms. Database Rights: Under Section 2(p) of the Copyright Ordinance 1962, curated databases of proprietary domain training data enjoy independent copyright protection as compilation literary works. --- Ethical and Regulatory Trajectory: What Tech Leaders Must Monitor The Ministry of Information Technology & Telecommunication (MoITT) and the Intellectual Property Organization of Pakistan (IPO-Pakistan) are actively drafting national artificial intelligence policy frameworks. Future statutory revisions are expected to address: Compulsory watermarking for commercial AI-generated imagery and video content; Explicit statutory recognition of computer-generated works along the lines of the UK Copyright, Designs and Patents Act 1988 (CDPA Section 9(3)), where the author is deemed to be the person who made the arrangements necessary for the creation of the work; Strict compliance mandates under the Prevention of Electronic Crimes Act 2016 (PECA) prohibiting deepfake generation, identity theft, and algorithmic consumer deception.