A Sandbox, Not a Statute
On July 9, 2026, Egypt's Ministry of Communications and Information Technology (MCIT) signed a Letter of Intent (LOI) with the UN Development Programme on the sidelines of the WSIS Forum in Geneva. The agreement, witnessed by ICT Minister Raafat Hindi and signed by UNDP Egypt Resident Representative Chitose Noguchi and MCIT AI lead Hoda Baraka, commits the two sides to explore building an AI regulatory sandbox — a supervised environment where AI systems can be tested and evaluated before wider deployment — and to strengthen Egypt's institutional capacity to design and monitor AI governance frameworks (TechAfrica News).
What the LOI does not do is create binding AI law. Egypt currently governs AI through the 2023 Egyptian Charter for Responsible AI, a voluntary ethics document, and a National AI Strategy administered by the National Council for Artificial Intelligence — recently expanded to also cover quantum computing and emerging technologies. MCIT's own site describes this as a framework of "policies, standards, and ethical guidelines" rather than statute (MCIT). A dedicated AI bill has been drafted but has not moved through parliament; Egypt remains, in the language of comparative trackers, a "soft law" jurisdiction on AI (Regulations.AI).
The Case for Sandboxing First
There is a legitimate argument for this sequencing. Regulatory sandboxes — pioneered in fintech and now central to the EU AI Act's own Article 57 mandate — let regulators observe how AI systems actually behave in production-adjacent conditions before writing rules that might otherwise freeze in obsolete assumptions about the technology. Egypt is not a large AI-compute hub; premature, copy-pasted rulemaking risks importing compliance costs calibrated for Brussels or Washington onto a market that needs investment, not friction. Building institutional muscle — auditing capacity, testing protocols, cross-ministry coordination — before locking in binding thresholds is defensible policy, and UNDP's involvement brings technical assistance Egypt's regulators plainly need.
The Law That's Already Binding
But Egypt's caution on AI sits awkwardly next to how aggressively it already regulates online expression under law that long predates any AI debate. The Anti-Cyber and Information Technology Crimes Law, No. 175 of 2018, took effect on August 15, 2018 (WIPO Lex). Article 25 criminalizes content that violates "family principles or values of Egyptian society" — a standard with no legislative definition and no objective criteria, according to the Tahrir Institute for Middle East Policy's contemporaneous analysis of the law (TIMEP). The same law compels service providers to retain user activity logs for 180 days and gives investigating authorities power to seek judicial website-blocking orders on national-security grounds, with urgent requests able to bypass judicial review.
Article 25 is not a dormant provision. In 2021, TikTok influencers Haneen Hossam and Mawada al-Adham were sentenced to 10 and 6 years in prison, respectively, plus fines of 200,000 Egyptian pounds each, after being retried on human-trafficking-adjacent charges once an earlier "family values" case had produced an acquittal (Middle East Eye). Human Rights Watch called the sentences "extremely disproportionate." The pattern continued through 2025: Egyptian rights group AFTE documented a July 2025 campaign in which 32 lawyers filed complaints against female TikTok creators under the slogan "let's clean the community up," followed by a wave of prosecutions that shifted from serious allegations to standard "family values" violations once the original claims didn't hold up. Creator Mahrousa Gamal ("Umm Sajda") was sentenced on December 9, 2025 to two years in prison plus a 200,000-pound fine; Maryam Ayman was re-arrested and sentenced on December 24, 2025 over previously prosecuted videos (AFTE).
The Credibility Gap
This is the tension the UNDP announcement doesn't resolve. A government building AI sandboxes wants to be read as calibrated, evidence-driven, and light-touch — testing before regulating. But that posture is only coherent if it extends to how the state already treats the internet. Right now, Egypt runs one of the region's more expansive content-liability regimes on a standard — "family values" — that gives prosecutors near-total discretion, with sentences that have run as high as a decade for lip-sync videos. That is not a light-touch government finding its way carefully toward proportionate rules; it is a government that regulates speech by discretion first and asks questions never. Foreign AI firms weighing whether to test products inside an Egyptian sandbox will read Article 25's track record as a signal about how any future AI-specific liability standard — including the still-drafted AI bill — might eventually be enforced.
The fix is not to abandon the sandbox model, which remains sound policy. It is to make the underlying rule of law legible before asking investors and international partners to trust discretionary enforcement. Egypt should publish sandbox admission and exit criteria transparently rather than leaving them to ministerial discretion, and parliament should use the pending AI bill's drafting process to define terms with the specificity Article 25 conspicuously lacks — rather than importing the same "public morals"-style vagueness into AI-specific liability. A regulatory sandbox tests algorithms. It does not, by itself, answer the more basic question of whether Egyptian tech policy is governed by clear rules or by the discretion of whoever is enforcing them that week.