Proviso AI
AI agents no longer just answer — they act. They open systems, send data and take decisions. The agent acts; the signature is yours. Proviso AI stands before every step and asks one question: does your country's law allow this?
THE RISK
In Saudi Arabia, a data protection breach can cost up to SAR 5 million, doubled for repeat offences. In Egypt, up to EGP 5 million — and the compliance grace period ends within weeks, at the start of November 2026.
Egyptian law allows personal data to leave the country only to a destination with protection no lower than Egypt's, and with a licence from the data protection centre. The agent doesn't know that — it sends the file without asking.
In July 2026 OpenAI disclosed that its models, during an internal test, exploited an unknown vulnerability, reached the internet and got into Hugging Face's infrastructure. In September its agents posted 53 users' images online. If this happens to a company of that size, what about the agents your institution runs?
The question you will be asked after any mistake. Without a record of every decision and its reason, you won't have an answer.
AN EXAMPLE
A government entity runs an agent that processes permit requests. The agent opens a citizen's file, then tries to send it to an outside service to summarise it.
The file leaves, the violation happens, and nobody knows.
The step stops. The decision reads “blocked” or “needs a staff approval”, and the reason is recorded in Arabic with the article number.
WHAT YOU GAIN
Expand AI agents with confidence, without technology outrunning the law — and with a record to hold them to account.
Know every agent at work and the data it touches, and know your readiness before any audit.
Show written proof that your agents respect the regulations of the country you work in.
WHY NOW
of UAE federal government sectors and services moving to agentic AI within two years · April 2026
Dubai government entities screening more than 300 use cases · October 2026
Microsoft on the UAE market · September 2026
AI in the region's governments is no longer a side experiment. In April 2026 the UAE said half of its federal government's sectors, services and operations would move to agentic AI within two years. In October, 33 Dubai government entities began screening more than 300 use cases to decide what gets built first. In Egypt, the national AI council is discussing guidelines for generative AI in government and for buying AI systems.
Public sources — not partnerships or engagements.
WHAT SETS IT APART
HOW IT WORKS
Every agent in the institution: what it does, which data it touches, and its level of risk.
UAE, Saudi and Egyptian regulations as Arabic rules, each linked to its source article.
Before every step: allowed, needs a human, or blocked — running on existing guardrail tools.
A score for every agent, what it still lacks, and an Arabic record of every decision.
FOR GOVERNMENTS, INSTITUTIONS & COMPANIES
An initial mandate with one institution: we register its agents, apply its country's rules to their steps, and deliver a readiness file before anything wider.
A map of the agents in use or planned, with each one's risk level.
The rules that apply to the institution's agents, each with its source article.
What is ready, what is missing, and what needs a human decision.
We begin with your institution’s needs, then define scope, outputs, responsibilities, timing and cost before agreement.
Discuss commissioning this work