Iran's intellectual property laws leave AI-generated art and inventions in legal limbo; revising statutes or creating tailored rights is needed to protect innovation without eroding human creators' interests.

Main Finding

Current Iranian intellectual property law has significant regulatory gaps for works autonomously generated by AI. The paper concludes that without targeted legal reform (either by assigning rights to specific human agents, creating a sui generis IP right for AI outputs, or otherwise clarifying ownership and liability), Iran — like most jurisdictions — faces uncertainty that will distort incentives, complicate enforcement, and raise ethical concerns. Comparative models (e.g., UK statutory allocation to the person who made the arrangements) offer possible templates, but each approach carries economic and policy trade-offs.

Key Points

• Problem framed: AI systems can autonomously generate texts, images, music, and inventions; traditional IP doctrines presuppose a human creator/inventor.
• Three principal legal approaches to ownership of AI outputs:
• Grant rights to the developer/owner of the AI system (aligns rewards with those who invested in and built the tool).
• Grant rights to the user/prompt-designer who provided inputs and exercised selection/curation (recognizes user creativity).
• Treat AI outputs as ownerless/public domain (no copyright/patent protection for fully autonomous outputs).
• Comparative practice:
  • United States: Copyright Office and courts have generally refused protection for non‑human-created works (e.g., extension of logic from Naruto/monkey selfie outcomes).
  • United Kingdom: CDPA s.9(3) recognizes as author the person who made the arrangements for a computer‑generated work (statutory assignment to a human agent).
  • EU and many jurisdictions: tacitly follow a human‑creator requirement; no uniform stance.
  • Patents/DABUS litigation: most patent offices (US, EU, UK) require a human inventor; exceptional rulings (South Africa granted, Australian lower court initially accepted but was overturned on appeal) illustrate divergence.
• Iranian legal framework:
  • No explicit AI-specific IP provision. Key statutes: 1969 Law for the Protection of Authors, Composers, and Artists’ Rights; 2000 software protection law; Patent/Industrial property laws (2007, replaced by Industrial Property Law in 1403/2024–25) — none squarely address AI‑authored outputs.
  • Software law can protect the AI system (code) and its developers, but does not clearly cover the independent artistic/creative outputs produced by the system.
  • Under Iranian civil principles, only persons (natural/legal) have legal capacity, complicating attribution to AI itself.
• Additional legal issues:
  • Moral rights: meaningless to attribute to AI; assigning them to humans raises ethical/authorship concerns.
  • Liability and infringement: training on copyrighted works raises potential infringement (e.g., Getty Images suits against image‑generating models); responsibility for infringing outputs (developer vs user) is unsettled.
  • Enforcement challenges: detecting infringement, apportioning responsibility, and defining the threshold of human creative contribution are difficult in practice.
• Policy recommendations from the paper:
  • Revise existing IP statutes or introduce tailored rules for AI outputs.
  • Options include statutory allocation to a human agent (user, developer, or funder), creation of a new sui generis right, or clear rules for moral rights and liability allocation.
  • Balance needed between promoting AI innovation and protecting human creativity and fairness.

Data & Methods

• Methodological approach: doctrinal legal research and comparative legal analysis.
  • Systematic literature review of academic sources, policy reports (e.g., WIPO), and legal scholarship on AI and IP.
  • Comparative review of statutory texts and administrative/court decisions in Iran, the United States, the United Kingdom, the EU, Australia, South Africa, and notable litigation (e.g., DABUS patent applications, Getty Images litigation, Colorado AI art contest, Naruto case).
  • Analytical synthesis of legal doctrines (copyright, patent, software protection), statutory interpretation, and practical enforcement problems.
• No empirical or econometric data; analysis is qualitative and normative, grounded in legal texts, case law, and secondary literature.

Implications for AI Economics

• Investment incentives and innovation:
  • Legal uncertainty reduces expected returns and increases risks for firms investing in creative AI systems; clear rules increase predictability and encourage R&D.
  • Granting rights to developers concentrates returns to technology producers, likely boosting upfront investment in model development but increasing market concentration (winner‑take‑most dynamics).
  • Granting rights to users/prompt designers may stimulate demand‑side experimentation and broaden the base of creators, creating markets for creative services and prompting more varied use cases.
  • A public‑domain approach maximizes downstream access and spillovers but may undermine private incentives to develop costly models and datasets.
• Market structure and competition:
  • Developer‑centric rights favor incumbent platforms that control models and training data, raising barriers to entry and potential for monopolistic pricing of generated content.
  • User‑centric or more permissive regimes lower vertical control by platform owners and can foster an ecosystem of intermediary services (prompt‑engineering, marketplaces for curated outputs).
  • Sui generis regimes could be designed to balance incentives (limited exclusivity or compulsory licensing) but add regulatory complexity and compliance costs.
• Labor and creative industries:
  • IP allocation affects how economic rents are distributed between AI firms, human creators (artists, writers, musicians), and end users; policies will influence substitution versus complementary relationships between humans and AI.
  • Strong exclusive rights to AI outputs may displace human creative labor; shared/limited rights and transparency requirements can help preserve markets for human creators.
• Data and training‑set markets:
  • Liability and clearance rules for training data will shape the emergence of licensing markets for large copyrighted corpora; strict liability or heavy compliance costs could raise model training costs and limit entrants.
  • Clear rules promoting data licensing and attribution markets could create new revenue streams for rights‑holders and lower litigation risk for model builders.
• Transaction and enforcement costs:
  • Ambiguous ownership and liability raise transaction costs (contracts, licensing negotiations, litigation) and slow commercialization.
  • Statutory clarity (e.g., UK‑style allocation or well‑specified sui generis rights) reduces transaction costs and enables efficient market contracting.
• Policy trade-offs (summary):
  • Developer ownership: pro‑R&D, pro‑concentration. Good for model investment; risks market power.
  • User ownership: pro‑broad creative participation, may require clear thresholds for authorship; encourages services around creative use.
  • Public domain: maximal access and innovation spillovers; weak private incentives for costly model training.
  • Sui generis with calibrated exclusivity or compulsory licensing: potential middle path but administratively complex.
• Practical policy recommendation for economists/policymakers:
  • Evaluate legal reforms not only on fairness or doctrinal fit but on dynamic economic effects: investment incentives, competition, labor impacts, and the creation of data licensing markets.
  • Consider hybrid approaches: e.g., allocate baseline economic rights to developers with mandatory licensing or revenue‑sharing mechanisms for downstream commercial use, or grant users limited exclusive rights when demonstrable creative contribution exists.
  • Promote transparency, provenance labeling, and data‑licensing infrastructure to reduce enforcement costs and enable functioning markets.