Saudi Arabia Saudi Arabia Anti-Cybercrime Law content moderation

Saudi Arabia's AI Media Labelling Principles Are Sensible, but Their Weight Depends on the Cybercrime Law Behind Them

Disclosure of AI-generated media is a proportionate idea. Saudi Arabia's vague Article 6 offences could turn a voluntary label into a liability.

Saudi AI Media Rules and Article 6 People of Internet Research · Saudi Arabia 8 Guiding principles in framework Transparency, accuracy and account… 5 years Article 6 maximum prison term Imprisonment not exceeding five ye… SR3M Article 6 maximum fine A fine not exceeding three million… 6 People referred to prosecution GAMR referral citing Article 6, pa… peopleofinternet.com
Saudi AI Media Rules and Article 6 People of Internet Research · Saudi Arabia 8 Guiding principles in fr… 5 years Article 6 maximum prison term SR3M Article 6 maximum fine 6 People referred to prosecution peopleofinternet.com

Key Takeaways

Saudi Arabia has set out national principles for the use of artificial intelligence in media. Arab News, citing the Saudi Press Agency, reported the document on 6 November 2026. It calls on "media practitioners and content creators to disclose when material has been generated or assisted by AI, using clear labels, watermarks or metadata," so audiences can tell synthetic content from authentic records. The text is built around eight principles, and the report lists them: transparency and disclosure; factual accuracy and credibility; privacy and protection of public figures; preventing misleading content; accountability across the content lifecycle; impartiality; public awareness; and ongoing impact assessment and regulatory compliance.

What we can and cannot confirm

Some facts about the document are still uncertain. The Arab News report names SDAIA and the General Authority of Media Regulation (GAMR) as organisers of the "AI for Impact" forum where the initiative was discussed. It does not say that they issued the framework. Other coverage says the Ministry of Media issued it together with GAMR, the Saudi Broadcasting Authority, SDAIA and the Saudi Media Forum. We could not retrieve the primary text. The report also does not say whether the principles are binding, and it does not mention the Anti-Cybercrime Law. The language it quotes ("calls on", "guidelines") reads as recommendatory. Readers should treat the binding status as unresolved.

The date also needs care. The framework was reported on 6 November 2026, but it is not new: it was launched at the Saudi Media Forum earlier in 2026, and SDAIA has issued related material before. Its Deepfakes Guidelines were published in May 2025 under the number SDAIA-P119 and updated in May 2026, according to Arab News. Those guidelines say content creators "must apply visible, tamper-resistant watermarks to all synthetic content." Any claim that this media framework is a 2024 document is not supported by what we could verify.

The strongest case for the framework

The case for the framework is a serious one. Synthetic audio and video are now cheap to produce. A fabricated clip of a public figure can spread faster than any correction. A disclosure norm costs honest publishers little, helps audiences, and gives platforms a standard signal to rely on. It also puts the burden on the party that knows how the content was made. A label states a fact about provenance and says nothing about the viewpoint. That is why disclosure is the least speech-restrictive tool regulators have for synthetic media. We would rather see it than outright bans on AI-generated content.

Where the proportionality problem lies

The difficulty is the law the principles sit beside. Article 6 of the Anti-Cyber Crime Law, issued by Royal Decree No. M/17 in 2007, applies to anyone who commits the cyber crimes it lists. The penalty is "imprisonment for a period not exceeding five years and a fine not exceeding three million riyals or to either punishment." The first listed offence is the "production, preparation, transmission, or storage of material impinging on public order, religious values, public morals, and privacy, through the information network or computers." The English text is hosted by WIPO Lex, which notes that the Arabic text governs. The statute is also published by the Ministry of Communications and IT.

The phrases "public order" and "public morals" have no stated limits in that wording. Regulators are actively using them. In November 2025, GAMR referred six people to the Public Prosecution for online content it said incited public opinion. Gulf News cited Article 6, paragraph 1, and reported the maximum of five years and SR3 million. These were referrals, not verdicts. GAMR said it "will not tolerate media content that threatens public order or breaches media regulations."

This matters for a labelling scheme. A disclosure rule is easy to follow when the penalty for non-compliance is a warning or a takedown. It is a different matter when the same regulator can also reach for criminal provisions with broad wording. Suppose a creator forgets a watermark, or a metadata tag is stripped by a platform. They could then face a regulator that already has a wide statute available. We found no evidence that this has happened, and we do not claim it has. The risk is structural. Compliance with a voluntary standard can become evidence in a content case, and an uncertain binding status makes that risk harder to price.

What a proportionate version looks like

Three design choices would keep the benefits of disclosure and reduce the speech risk.

SDAIA's own framing in its deepfakes work points the same way. Arab News reports that the guidelines say penalties for misuse should be "proportional to the severity, intent, and recurrence of violations." Applying that principle to media labelling would be a sound default.

The innovation stake

Saudi Arabia has put AI at the centre of its economic strategy, so the clarity of these rules matters to the sector. Newsrooms, studios and creator-tool companies need to know what they must disclose and what they will be held liable for. Clear, narrow duties would encourage investment in provenance tools. A mix of soft principles and open-ended criminal law would encourage over-labelling at best and self-censorship at worst.

The disclosure idea is good. Its success will depend on the surrounding legal environment, and in particular on whether the Kingdom commits to keeping content offences and AI labelling apart. Until the binding status is published, publishers should label their AI use and keep records of how their content was made. Regulators should say publicly that a labelling lapse is not a cybercrime.

Sources & Citations

  1. Arab News Pakistan: Saudi Arabia sets out AI principles for media
  2. WIPO Lex: Saudi Anti-Cyber Crime Law (English translation)
  3. Ministry of Communications and IT: Anti-Cyber Crime Law (PDF)
  4. Gulf News: Saudi Arabia refers six people to prosecution
  5. Arab News: SDAIA deepfakes guidelines