Argentina’s president, Javier Milei, proposed this month to create a new class of company owned and operated by artificial intelligence, with human shareholders permitted but not required. The plan, set out in the Financial Times on 4 June and backed by a bill that would replace corporate legislation in force since 1972, was widely received as a leap into the unknown. In legal terms, it is not. The structure he describes already exists.
Since 2021, Wyoming has allowed a limited liability company to be managed by code rather than by people. Under the state’s decentralised autonomous organisation statute, software can make a company’s binding decisions; where that code and the written operating agreement conflict, the code prevails. The state imposes no requirement that a member be human. An artificial agent can, in practice, hold and direct such a company today.
Legal personhood is not a judgement about consciousness but an administrative device – long extended to companies, ships and trusts – that lets an entity own assets, enter contracts and be sued. Several jurisdictions have weighed formalising it for AI and drawn back. The European Parliament endorsed “electronic personhood” in 2017 by 396 votes to 123, then dropped the idea within two years after more than 150 specialists warned against it. Britain’s Law Commission revived the question in 2025 but recommended no change. Ohio has moved to prohibit it outright.
The unresolved problems are practical. A company governed by software cannot be imprisoned and treats fines as a cost, removing the deterrents that constrain human directors. Limited liability without a responsible person behind it is an efficient means of placing decisions beyond legal reach. The pressing questions are no longer philosophical: who is accountable when such a company causes harm, and whose law applies when it is incorporated in one country and operates in another.