Saudi Arabia’s New Copyright Regulations: What AI Companies Should Know

The Implementing Regulations

On 31 July 2026, Saudi Arabia published the Implementing Regulations of the Copyright Law in the official gazette (Umm Al-Qura). Issued under Royal Decree No. (M/169), the Regulations supplement the updated Copyright Law, which entered into force on 12 August 2026. Notably, they set out a detailed framework governing the use of copyrighted works in AI development in the Kingdom. The Regulations took effect upon publication.

The AI Development Exception

Article 26(4) of the Copyright Law permits the use of copyrighted works, without the author’s permission or payment of compensation, for the purpose of developing AI products and algorithms. The exception applies when the work has been lawfully published, the original copy has been lawfully acquired, and any copying is limited to what is necessary for that purpose.

Article 30 of the Regulations builds on this exception by setting out additional conditions for the copy and analysis of copyrighted works in connection with the development of AI algorithms and products, including that:

• Any copying must be limited to what is necessary for AI development and may not extend to republication, distribution, or direct commercial exploitation of the underlying work.
• Use of a copyrighted work in a purely commercial context is not permitted unless (a) the use is non-substantial in relation to the work or (b) it does not affect the work’s normal exploitation — satisfying either limb is sufficient. The scope of this restriction remains unclear and may require further guidance from the Saudi Authority for Intellectual Property (SAIP). One possible interpretation is that it targets situations where AI-generated output reproduces substantial portions of copyrighted material, which is used for commercial purposes. A broader reading, however, could bar commercial AI developers from taking advantage of the AI training exception, though that seems contrary to the purposes of the law. 
• Developers must maintain records of works used, including the type of work, source, purpose, and date of use, and provide those to any competent authority upon request. However, the Regulations do not specify the level of detail required, including whether records must be maintained for individual works, datasets, or on an aggregated basis. Further, the Regulations are silent on retention periods, format requirements, and standard of detail. Further guidance from SAIP may be needed.
• The use must not cause unjustified harm to the author’s legitimate interests or affect the author’s opportunity to exploit or derive financial benefit from the work.
• The Regulations prohibit adapting, republishing, making available to the public, or incorporating copyrighted works into final products, if doing so is unnecessary, unless the rights holder has consented or the work is in the public domain. The Regulations suggest that copyrighted material should not be incorporated into AI-generated output, however they do not address how this restriction applies to models. For example, it is unclear whether training data that is not retained in recognisable form in a deployed model constitutes “inclusion.” This remains an open question and may require further guidance from SAIP.

There are several aspects of this regulation that may require further guidance from SAIP to help developers of AI products navigate the Article 30 exception, including the scope of the restriction on causing unjustified harm to legitimate interests, the scope of the “purely commercial” use restriction, the required granularity of record-keeping, and the meaning of “inclusion in final products”, including how necessity is assessed.

Compulsory Licensing and Arabic-Language LLMs

Separately, the Regulations establish a compulsory licensing regime (Articles 31–43) under which Saudi nationals or entities may apply for licences to translate works into Arabic if the rights holder has not done so within prescribed timeframes. For translation, the timeframe is usually three years. Applicants must demonstrate reasonable efforts to obtain a contractual licence, pay fair compensation, and restrict distribution to within the Kingdom.

Although the regime is not explicitly aimed at AI development, it could increase the availability of lawfully translated Arabic-language content. This may be particularly relevant to AI companies that are developing models with Arabic-language capabilities.

Ownership of Copyright in AI-generated Works

Saudi Arabia’s draft intellectual property law contemplated specific rules for AI-generated intellectual property. Notably, a chapter on “Intellectual Property associated with Artificial Intelligence and Emerging Technologies” provided that intellectual property created using AI technology would be eligible for protection if a natural person made a meaningful contribution to its creation. By contrast, works where human contribution was insignificant or those generated independently by AI would fall into the public domain. For more detail on the draft law, see here.

The new Copyright Law and Implementing Regulations do not address ownership of copyright in AI-generated works. As a result, further guidance from SAIP may be required. This is particularly notable given that SAIP’s guidelines on AI-related patent applications confirm that inventions created with human contribution, using AI as an assistive tool, may be eligible for patent protection. For more information, see the relevant Guidelines.

Practical Implications

The combination of the Article 30 AI training exception and the compulsory Arabic translation licensing framework may enhance Saudi Arabia’s appeal as a jurisdiction for AI development. However, important questions regarding the scope and operation of the framework remain, requiring further guidance from SAIP. We will continue to monitor developments from SAIP and the Saudi Authority for Data and Artificial Intelligence (SDAIA). In the meantime, organisations should consider:

• Whether establishing or expanding AI training infrastructure in the Kingdom could allow them to benefit from the Article 30 exception.
• How the record-keeping and purpose limitation requirements will interact with their existing data governance frameworks.
• The scope of the “insubstantial use” and “normal exploitation” standards in Article 30, which are likely to generate interpretive disputes as the framework is applied in practice.
• The compulsory licensing pathway as a means of expanding lawful access to Arabic-language content for LLM training.

The authors would like to thank Norah AlKhalaf and Sitah AlJoaid for their contribution to this Article.