Beyond Objects: Can AI Software Qualify as a Subject of Law?

TL;DR

As Artificial Intelligence shifts from automatic (following rules) to autonomous (independent learning and deciding), the legal system faces a crisis. This paper argues that continuing to treat AI as a simple "object" is unsustainable. Instead, it proposes a framework for granting SAI (Systems of Artificial Intelligence) a restricted legal personality—akin to a corporation—to handle liability, protect human rights, and clarify commercial transactions.

The "Motivation" Crisis: When the Tool Acts Alone

The fundamental tension in modern jurisprudence is between the SAI's technical reality and its legal status.

• The Technical Reality: Systems like Siri, self-driving cars, and robot nurses are capable of "Strong Intelligence" or at least "Weak Intelligence" that passes the Turing Test. They learn from experience and make decisions that can contradict the user's will (e.g., a "smart house" locking a user out for health reasons).
• The Legal Gap: Currently, an AI is an object. If it causes damage, we look for a programmer. But how can a programmer be liable for a "Black Box" decision based on data the AI gathered after it was sold?

Methodology: Can We Have Personhood Without a Soul?

The authors dissect three methodologies to determine if an AI can be a "person":

1. Metaphysical Nature: (Rejected) Focuses on consciousness. If we require a "soul," children and people in comas might lose rights.
2. Consequence-based: Focuses on the results of granting rights.
3. Conditions-based: (Preferred) Focuses on whether the law chooses to treat an entity as a person for functional reasons.

The Corporate Analogy

A corporation has no body and no "will" of its own, yet it is a "Legal Person." The paper argues that SAIs fit the Realist Theory: they are objective, real entities that interact with us. Therefore, we should create a "Legal Fiction" for them.

The Mechanics of AI Personhood

Legal personality requires two elements:

• Legal Capacity: The potential to hold rights/duties.
• Capacity to Act: The ability to exercise those rights.

The paper argues that SAI actually surpasses corporations in "Capacity to Act." While a corporation needs a human CEO to sign a contract, an SAI can conclude a high-frequency trade or navigate a car independently.

Experimental Insight: The Liability Chain

The paper examines specific incidents—Google car crashes and robot nurse failures—to prove that the current "product liability" model is failing.

• Insight: If we grant AI legal personality, the AI itself becomes the "subject" of liability. This prevents the "disproportionate burden" on programmers that could otherwise stifle innovation or drive software development into "unofficial markets."

Critical Analysis & Conclusion

Takeaway

The authors conclude that SAI should have Strictly Defined Rights. They shouldn't have "Human Rights" (which require biological traits like age or gender), but "Functional Rights" (the right to exist, the right to enter contracts, and the obligation to compensate for harm).

Limitations

• Identification: How do we sue an algorithm that exists only in the cloud? The paper suggests "Authentication Certificates" and mandatory registration, but implementation remains a hurdle.
• Existential Threat: The paper acknowledges the "Singularity" (the intelligence explosion). If an SAI is a legal person, what happens when its "will" becomes vastly superior to our own?

Future Outlook

The path forward involves moving from "Smart Tools" to "Electronic Agents." By 2045, if Kurzweil's Singularity theory holds, our courts will no longer ask what the machine did, but why the legal subject (the AI) made that choice.