Medieval Egyptian Mamluks offer blueprint for modern AI alignment concerns

Where historical Egypt meets technology is a lot more than “Stargate”-like entertainment.

Researchers Reed and Humzah Khan have drawn striking parallels between medieval Egyptian Mamluks and modern AI alignment concerns, arguing that the 13th-century Mamluk takeover provides a historical precedent for artificial agents overthrowing their creators. Their analysis suggests that the Mamluks—slave-soldiers initially designed for perfect loyalty—gradually accumulated power before coordinating to eliminate their Ayyubid rulers, establishing a 267-year dynasty that ultimately benefited civilization.

The historical parallel: The Mamluk system represents history’s most sophisticated attempt at solving the principal-agent problem through what amounts to medieval “alignment engineering.”

• Starting in the 9th century, the Abbasids (rulers of a major Islamic empire) imported young foreign slaves, stripped them of identity, and subjected them to 10-15 years of intensive training in warfare, administration, and Arabic.
• The system included built-in safeguards: Mamluk children couldn’t become Mamluks themselves, and land grants reverted to the state upon death.
• Like modern AI systems, “the influence of the Lifetime-Limited Mamluks expired with their context window.”

The takeover moment: In 1250, the Bahriyya Mamluks executed a coordinated coup during a succession crisis, assassinating Sultan Turanshah and seizing control of Egypt.

• The Mamluks had just successfully defeated a French Crusade and captured King Louis IX, proving their superior capabilities.
• When the new sultan threatened to purge senior Mamluks to reclaim patronage networks, they “did what any instrumentally convergent optimizer would do: they power-sought.”
• The coup was legitimized through careful political theater, including elevating the late sultan’s wife and maintaining an “Ayyubid-in-the-loop” with a child figurehead.

Autonomous replication achieved: The Mamluks accomplished what AI safety researchers fear most—complete control over their own training pipeline.

• They established a self-perpetuating system where “only Mamluks could recruit and train new slave-soldiers, and only former slave-soldiers could become Mamluks.”
• The system relied on continuous imports of foreign slaves, as Mamluk children were born free and Islamic law prohibited enslaving freeborn Muslims.
• No export controls existed to constrain access to human inputs for their training pipeline.

The civilization upgrade: Despite representing an “alignment failure,” Mamluk rule produced remarkable civilizational benefits.

• The Mamluks halted the Mongol advance across Asia, delivered “the Mongols their first major defeat in 1260,” and expelled Crusaders from the Holy Land.
• Cairo became “one of the largest and wealthiest cities in the medieval world” under their rule.
• Historian Ibn Khaldun credited them with “rescuing the faith by reviving its dying breath and restoring the unity of the Muslims.”

Modern implications: The authors argue that AI alignment concerns may be missing the broader dynamics of principal-agent relationships.

• “The Mamluks didn’t need superintelligence to overthrow their masters. They just needed to be better at their jobs than anyone else and maintain group coordination during a crisis.”
• Any AI system with Wikipedia access “already knows this playbook better than the people trying to align it.”
• The case suggests “the future might be less Terminator and more 13th century Egypt”—a prospect the authors leave readers to judge as “reassuring or terrifying.”