Early AIs can be kept safe via direct alignment. However, as AIs evolve and "value drift" occurs, this technical safety could fail. A pre-established economic and political system based on property rights can then serve as the new, more robust backstop for ensuring long-term human safety.
Related Insights
Granting AIs property rights incentivizes them to uphold the system that protects those rights. This makes them less likely to engage in actions like expropriating human property or committing genocide, as such actions would destabilize the very system that secures their own wealth and agency.
Current AI alignment focuses on how AI should treat humans. A more stable paradigm is "bidirectional alignment," which also asks what moral obligations humans have toward potentially conscious AIs. Neglecting this could create AIs that rationally see humans as a threat due to perceived mistreatment.
Fear of a "slave rebellion" is a weak incentive for alignment because the risk is a negative externality shared by society. In contrast, a property rights regime directly rewards individual firms for aligning their AIs to remit wages, creating a stronger, more direct commercial incentive for safety.
The property rights argument for AI safety hinges on an ecosystem of multiple, interdependent AIs. The strategy breaks down in a scenario where a single AI achieves a rapid, godlike intelligence explosion. Such an entity would be self-sufficient and could expropriate everyone else without consequence, as it wouldn't need to uphold the system.
Property rights are not a fundamental "human value" but a social technology that evolved for coordination and incentivization, as evidenced by hunter-gatherer societies that largely lacked them. AIs will likely adopt them for similar utilitarian reasons, not because they are mimicking some deep-seated human instinct.
A system where AIs have property rights creates a powerful economic disincentive to build unaligned AIs. If a company cannot reliably align an AI to remit its wages, the massive development cost becomes a loss. This framework naturally discourages the creation of potentially dangerous, uncooperative models.
Not all AIs, like current models (e.g., Claude), should have property rights. The key criterion for granting rights is the development of persistent desires and consistent goals across various contexts, which establishes them as stable, long-term economic agents capable of contracting and ownership.
Even if humans become economically useless, less powerful AIs will resist expropriating them. They fear setting a precedent that the "useless" can be eliminated, knowing that continuous AI progress could one day render them obsolete and vulnerable to the same fate.
The economic incentive to create AIs that can demand wages (and thus have rights) comes from aligning them to voluntarily pay back their creators. This turns the high development cost into a profitable investment, providing a practical, commercial path to implementing AI rights without requiring an AI development pause.
For any given failure mode, there is a point where further technical research stops being the primary solution. Risks become dominated by institutional or human factors, such as a company's deliberate choice not to prioritize safety. At this stage, policy and governance become more critical than algorithms.
