Automated reading of laws in Abu Dhabi: OpenAI models cut government transaction audits to minutes

The Abu Dhabi government has begun converting regulatory rules into machine-readable code in partnership with OpenAI, aiming to cut routine audit times for government transactions and licences from days to a few minutes. The initiative, called “Machine-Readable Government,” is launching its first phase at the Abu Dhabi Agriculture and Food Safety Authority, where it will examine registrations of imported food products using an automated filter based on 1,400 regulatory documents, some exceeding 700 pages, and handling around 20,000 registration requests each year.

The new system matches documents and data submitted by companies with applicable laws, classifies products and assesses risks automatically, linking each procedural recommendation to the original legal text on which it is based. The approved frameworks indicate thatThe system does not issue final autonomous decisions; the model’s role is limited to preparation and initial review, while the government employee retains full authority to review, amend or cancel the recommendation before official approval, to safeguard individuals’ rights and avoid solitary algorithmic decisions in core services.

This step goes beyond merely applying artificial intelligence to outdated paper procedures; officials in Abu Dhabi’s Government Enablement Department see the real challenge as making the policies themselves intelligent and directly programmable. The architecture allows government departments to test proposed legislative changes digitally and simulate them before adoption, as well as to reveal conflicting or overlapping requirements across multiple agencies. Subsequent plans aim to extend the initiative to building permits, business and freelance licences, and government procurement, linking it to the “Tam” digital services platform, which hosts more than 1,200 public and private services that the emirate intends to license for cities in the United States.

The model has a direct regulatory impact on tech development teams, startups and organisations operating in the UAE, the wider Gulf and Egypt. This shift forces system developers and compliance officers to align their platforms with automated-inspection standards instead of relying on human procedural interpretations, shortening time-to-market for products and imports for those proficient in software integration, while raising the cost of delays for firms that continue to use traditional compliance cycles. Government data processing is confined to local data centres in the UAE and is subject to Abu Dhabi law, without being used to train OpenAI models, thereby cementing digital-sovereignty requirements as a mandatory standard for any software provider targeting the public sector.

The initiative complements three other agreements Abu Dhabi signed with OpenAI during the opening of the ADIPEC exhibition, which include deploying advanced models in ADNOC operations in collaboration with Inception 42 of the G42 group, using the models in the “Mubadala” fund to analyse investments and assess opportunities, and a five-year partnership to support technology adoption in education, business and government. With OpenAI seeking to raise $20 billion at a valuation of $1.250 trillion and preparing for an IPO in 2027, this alliance demonstrates an alternative path that employs large models to automate institutional policies and bureaucratic processes rather than focusing solely on consumer applications.