Argentina’s government has decided to push the boundary between automation and legal autonomy by proposing a bill that would allow entities run entirely by artificial intelligence or robots to exist as legal subjects. The draft, sent to Congress, introduces the “non-human corporation” — a novel legal construct capable of signing contracts, holding assets, and operating on the market without any human being in charge. Yet the same proposal contains clauses acknowledging that human intervention cannot be completely eliminated, a detail that is as technical as it is political.

What the bill actually says

The new legal category aims to grant limited personhood to AI systems and robots that manage commercial operations. In theory, such an entity could conduct negotiations, manage assets, and even face tax obligations without a human director. Born in a context of strong regulatory innovation, the project nevertheless suggests that critical functions like legal representation, liability for damages, and ultimate decision-making oversight still require human figures or external supervision mechanisms.

Why this matters to those building AI infrastructure today

The news may read like legislative reporting, but for those developing and managing enterprise AI systems it sends a clear signal: delegating legal acts to a software agent immediately raises the question of who controls the infrastructure running that agent. If an algorithm signs contracts fully autonomously, any organization deploying it will need to transparently show where and how those decisions are made — and can only do so if the entire operational chain is traceable.

In this light, on-premise or self-hosted deployments take on a different weight compared to generic cloud solutions. Those adopting local stacks can precisely define audit policies, data residency, access levels, and log sessions — all elements that become essential when a system is asked to perform acts with legal value. It’s no coincidence that heavily regulated sectors like finance and healthcare are already evaluating on-prem architectures for their LLMs and autonomous agents.

The Argentine case, despite its specificity, replays a common theme for every agentic AI initiative: the trade-off between efficiency and accountability. As language models and robots become more capable, the question isn’t whether they can operate alone, but whether the technical and legal framework surrounding them can bear the consequences. For those tracking this evolution, AI-RADAR offers analytical frameworks that help assess deployment constraints — from data sovereignty to pipeline transparency — without pushing a predefined technology choice.

The Argentine proposal may remain a theoretical exercise or become a global precedent. Either way, it reminds industry practitioners that the next generation of autonomous systems will demand infrastructure designed not just to scale, but to answer for its actions.