Autonomous Legal Entities are Already Possible Under American Law

If you ask a hundred lawyers whether a software system or a robot can buy a house or file a lawsuit, all of them would be likely to answer ‘no.’  But because of the extreme flexibility of limited liability companies (LLCs) under US law, software and other artificial systems in fact can get basic legal personhood or at least a very close surrogate of it.  That is, they can enter contracts, own property, sue, and be sued. 

The modern American LLC is an amazingly flexible structure.  At its core, it is a legally recognized entity controlled however its organizers want it to be controlled.  It is flexible enough to support a governance structure involving a single person (the very common ‘single-member LLC’) or one as complex as the United States government (with, say, a bicameral board, an executive of any level of intelligence, and an independent group of adjudicators who resolve internal disputes).  Conventional businesspeople take advantage of this flexibility to suit their circumstances.

Nothing in American statutes requires that LLCs remain conventional, however.  LLC law is flexible enough to permit software or any autonomous system whose output or operating state can be proven to a court to have direct legal effect.  For example, an operating agreement may condition rights or duties on the output of a software program.  In some respects, this is already commonplace; for example, business owners may be paid according to complex formulas executed entirely by software, with no human understanding the precise details of the calculation.  The calculation is nonetheless enforceable if the operating agreement makes it so.

Members of an LLC can amend an operating agreement, so the ability for software to have legal effect doesn’t, by itself, mean that software can act autonomously or assume the functional rights of legal persons.  In at least many American jurisdictions, however, the LLC is flexible enough to enter a durable state in which it has no members:  the single-member LLC can become what I’ve called a ‘zero-member LLC.’  The mechanics of this are surprisingly straightforward; just as Douglas Adams described flying as falling and missing the ground, an LLC can have zero members if all the members simply dissociate under conditions that do not cause the LLC to dissolve and terminate.

As an example, under the Revised Uniform Limited Liability Company Act (RULLCA), which is the basis for more than twenty US states’ LLC laws, the default rule is that an LLC dissolves upon ‘the passage of 90 consecutive days during which the company has no members.’  But as I have explained in detail in several articles, this is just a default rule, and under RULLCA an LLC’s operating agreement is permitted to change or eliminate it.  Perhaps more evocatively, New York’s LLC statute explicitly permits an LLC with no members to exist for a period of ‘one hundred eighty days or such other period as is provided  for  in  the  operating   agreement  after  the  occurrence  of  the  event  that  terminated  the continued membership of the last remaining member.’  Once an LLC has no members, software is free to govern the entity without interference from pesky humans.

As I describe in a recent article called ‘Are Autonomous Entities Possible?’ lawyers who are told of this statutory logic tend to resist it.  For example, a typical response is that courts will find ways to declare such a result to be ‘absurd’ or to hold that the true intent of the statute must not have been to cede such importantly human characteristics to autonomous operating agreements that delegate their decisions to software.  As I argue in that article, however, I think such objections are misplaced. For one thing, most simply, the policy of business-organization statutes in the US is moving rapidly toward, rather than away from, giving legal scope and authority to software.  Vermont, for example, recently passed a statute authorizing a new type of LLC that ‘may provide for its governance, in whole or in part, through blockchain technology’ (emphasis added).

It is also unclear how common-law courts would stop my techniques from operating even if they were motivated to.  Suppose a homeowner sells her house to an LLC operated by a robot.  Nothing in LLC law requires the LLC to disclose its operating agreement as the precondition to such a sale, so no question is likely to arise immediately.  If a dispute about the LLC’s permissibility arises later, what is there to challenge, precisely, and who is able to challenge it?  Should the homeowner be able to sue to recover both her home and the money she paid just because she’s offended by the idea that an entity controlled by software bought it?

Even if statutory policy were opposed to autonomous entities and even if courts were institutionally suited to police them, it would be very difficult to draw lines between impermissibly ‘autonomous’ and permissibly ‘nearly autonomous’ entities.   There is no fundamental difference between (on one hand) a conventional business controlled by procedures involving many humans and (on the other) a novel business controlled by software set in force by some humans and affecting others.  From this perspective, conventional business organizations may simply be a quaint form of artificial intelligence.  Because everyone agrees that single-member LLCs are permissible under US law, it would take only one cooperating human being to authorize functionally autonomous software anyway; that one person has some sort of formalistic authority to change an operating agreement doesn’t seem to make much of a difference when evaluating the viability of an entity.  And what if that single authorizing human is restricted by contract from changing the agreement?  To put it differently, is there really a difference between a zero-member LLC and an LLC with thousands of members who have no power because the operating agreement says they must act unanimously, if at all?

Finally, because of the internal affairs doctrine, if one jurisdiction permits autonomous entities, they are likely to be able to operate in others. So autonomous entities are already out of the bag, and it would take nothing short of a fundamental overhaul of business law to prevent them.