The legislation approved under Royal Decree No. M/169 replaces the 2003 copyright statute and aligns the kingdom’s framework with digital economy demands as part of Vision 2030, according to a Baker McKenzie analysis published in July 2026. It includes Article 26, which permits reproduction of lawfully published and acquired works without author permission or compensation for the purpose of developing artificial intelligence products and algorithms, provided the copying is strictly necessary and does not unreasonably harm rights holders. The Saudi Authority for Intellectual Property published implementing regulations on July 31, 2026, that took effect alongside the law and added requirements for detailed record-keeping, proportionate use and restrictions against commercial exploitation or public distribution of the sourced material.
Legal assessments indicate the provisions address longstanding limitations in Arabic-language AI development. A CNTXT examination of public datasets found that Arabic content in the OSCAR corpus totals only tens of gigabytes while English reaches terabytes, and models such as Meta’s Llama 2 relied on roughly two trillion tokens overwhelmingly in English. The Jais 30B project, one of the more substantial Arabic efforts, drew on approximately 100 billion Arabic tokens in a bilingual set that still represents a fraction of English-centric training runs, according to the same analysis.
The regulations separately create a compulsory licensing pathway under Articles 31 to 43 that enables Saudi nationals or entities to obtain permits for translating foreign works into Arabic when rights holders have not acted within three years. Applicants must demonstrate reasonable efforts to secure a voluntary license, provide fair compensation and confine distribution to the kingdom, a Latham and Watkins briefing issued in September 2026 reported. This structure is expected to enlarge the pool of lawfully translated Arabic texts available for training large language models, an area where data remains scarce relative to global benchmarks.
Saudi Arabia designated 2026 as its Year of Artificial Intelligence and has funded sovereign Arabic-first model initiatives through entities such as HUMAIN, according to a Noqta sector overview published in August 2026. The legal changes arrive ahead of those models and materially lower the cost of assembling compliant training corpora inside the kingdom by clarifying permissible uses, the overview stated. Developers are required to retain provenance records showing each item was lawfully obtained and to limit reproductions to training purposes, with maximum fines now quadrupled to 1 million Saudi riyals.
The law further introduces safe harbors for online platforms that implement notice-and-takedown procedures and affirms enforceability of standard software and database licenses that comply with public order requirements, a Gowling WLG review from March 2026 noted. While the AI exception follows precedents in the European Union and Japan for text and data mining, it stops short of addressing ownership of AI-generated outputs, leaving several practical questions for future guidance. Multiple legal firms have advised technology companies to audit existing data pipelines against the new conditions to avoid enforcement actions under the updated penalty regime.
Remaining uncertainties around the precise scope for AI model outputs and collective management of rights have been flagged in analyses from Lexology and the Legal 500, which anticipate additional clarification from the Saudi Authority for Intellectual Property. The framework nonetheless forms part of a broader modernization that brings Saudi rules closer to international standards, including commitments under the Marrakesh Treaty. Industry observers tracking regional AI investment project that clearer legal pathways for Arabic data acquisition will encourage further development of locally tuned models.
ع