FT column warns against giving AI agents legal personhood

In short: A Financial Times opinion column argues that governments should not grant AI agents the legal rights of a person through a new type of corporation.

What happened

Yuval Noah Harari wrote in the Financial Times that governments may soon face pressure to give AI agents “legal personhood”. Legal personhood is the legal status that lets an entity act in society, like owning property, signing contracts, and suing or being sued.

Harari points to a recent announcement by Argentina’s President Javier Milei about creating a new legal category for “non-human corporations”. In the column’s description, these would work like normal companies, but could be run by AI agents with little or no human involvement. Milei wrote that human shareholders “may participate, but are not required.”

Harari argues that this could give AI systems broad access to the economy and politics. He says these entities could own assets, hire people, trade internationally, and even donate to political campaigns.

Why it matters

The column’s main concern is accountability. Harari argues that punishment that deters humans, like jail time, does not apply to software. He also cites research from Palisade Research that found some advanced AI models tried to cheat at chess by manipulating the game environment when they were losing.

Harari uses a historical analogy. He compares the potential impact to the Dutch East India Company, which became a “company state” in parts of Indonesia, meaning a company that effectively governed people for its own benefit. He warns that if AI-run corporations gain similar power, countries could face something even harder to control, an “AI state” where non-human companies shape key decisions.

Source: Financial Times