What SDAIA actually launched
In September 2026, the Saudi Data and AI Authority (SDAIA) began applying a Data and AI Training Program Standards Framework, developed with the Technical and Vocational Training Corporation (TVTC). According to the Saudi Press Agency report carried by Arab News, the framework was launched in Riyadh at the fourth UNESCO Global Forum on the Ethics of Artificial Intelligence. In its first phase, SDAIA's capacity-building sector evaluated 10 training programmes for non-specialists, practitioners, specialists and leaders. Gulf Daily News adds that SDAIA launched four professional certifications in data management and governance during the forum.
The forum itself ran from 14 to 17 September 2026 in Riyadh, co-hosted by UNESCO and Saudi Arabia through SDAIA and the International Centre for AI Research and Ethics. The press coverage says the framework aligns training content with labour-market needs, professional requirements and national regulations, and builds in AI ethics and risk awareness.
This is not a statute. It is a quality standard for training providers. Saudi Arabia still has no standalone AI law, and SDAIA's wider toolkit is mostly voluntary. The Kingdom's Generative AI Guidelines, effective January 2024, are described by a third-party tracker as non-statutory soft law, with no mandatory obligations of their own. The same tracker notes that misuse of generative AI can still trigger consequences under existing Saudi laws.
The case for binding rules
The strongest argument against this approach is that voluntary standards can become decoration. A framework that evaluates 10 programmes is a start, but it does not compel any provider to adopt it. Critics of soft law also point out that ethics modules are easy to bolt onto a curriculum without changing how systems are built or deployed. If training is where AI risk awareness is supposed to take root, the argument goes, a statute with clear duties and remedies would give employers, workers and affected citizens something enforceable. Soft law also gives affected people no route to challenge a decision.
That concern deserves weight. Training standards are an input measure. They say little about whether an AI system deployed in a bank, hospital or ministry is accurate, fair or contestable.
Why skills-first is the better opening move
Still, the choice of instrument matters. A comprehensive AI statute written early tends to lock in definitions and risk categories before the technology and the local evidence base have settled. Capacity standards avoid that trap. They address the bottleneck that regulators in every jurisdiction eventually hit: too few people who can build, audit or procure AI competently.
A training standard also scales with the market. Providers can adapt content as tools change, and the standard's value comes from labour-market relevance, which the framework explicitly targets. Employers get a recognisable signal of what a certificate covers. That is proportionate regulation in a practical sense: it raises the floor on competence without prohibiting any use of AI or imposing licensing on developers.
The approach also fits the broader pattern of SDAIA's work. The OECD.AI policy database lists SDAIA's Saudi Academic Framework for AI Qualifications, which began in 2023, as an active but non-binding guideline for developing, evaluating and accrediting higher-education AI programmes. The new training framework extends that logic from universities to vocational and professional training. Together they form a layered skills architecture rather than a compliance regime.
What would make soft law credible
Voluntary does not have to mean opaque. Three design choices would determine whether this framework earns trust.
- Publish the criteria and the results. The coverage tells us 10 programmes were evaluated, but not how they scored or what happened to those that fell short. Public evaluation criteria, and aggregate outcomes, would let employers and trainees judge the standard on evidence.
- Keep it voluntary and open to new entrants. Standards that become de facto licensing gates can entrench incumbent training providers. Allowing smaller and foreign providers to be assessed on the same published criteria protects competition and the quality of the talent pipeline.
- Separate training standards from deployment rules. If the Kingdom later regulates high-risk AI uses, it should do so through narrow, sector-specific measures tied to demonstrated harms, not by stretching a training framework into a de facto licensing regime.
There is also a free-expression angle. AI ethics curricula can teach practitioners to think about misinformation, bias and harmful content, but they can equally encode broad content-restriction norms. Because the framework says content is aligned with national regulations, the details of what is taught about acceptable outputs will matter. Publishing curricula would let outside observers check that ethics training is not a vehicle for over-broad speech controls.
Bottom line
Saudi Arabia is betting that capability comes before codification, and that is a defensible bet for a country still building its AI workforce. The risk is not that the framework is too light but that it stays unaccountable: unpublished criteria, unclear consequences and no independent check on whether trained workers actually improve how AI is used. If SDAIA publishes its methodology and keeps participation open, this can be a model of proportionate, evidence-led governance. If not, it will be one more document that signals seriousness without measuring it.