# Propositions from Letter: What Happens Once You Give AI Agents Legal Identity

**Citation:** Yonathan A. Arbel, Simon Goldstein & Peter Salib, Letter: What Happens Once You Give AI Agents Legal Identity, Financial Times (2026).

**Source:** [unpaginated Financial Times letter; human-curated repository summary reviewed because publisher body was inaccessible](https://www.ft.com/content/9451b5e9-035e-46e4-bafd-ede04dd13d8b)

**Review status:** 6 model-drafted, source-checked; 0 human-reviewed. This online-only work uses section-level unpaginated anchors.

## 1. The letter answers Yuval Noah Harari’s objection that legal identity for AI agents would dangerously empower machines

**Location:** Letter Summary, Response to Harari, unpaginated online source

Professors Yonathan A. Arbel, Simon Goldstein, and Peter Salib claim, in “Letter: What Happens Once You Give AI Agents Legal Identity,” an unpaginated Financial Times letter, under the repository’s “Letter Summary,” that Harari’s objection treats AI legal personhood as a dangerous grant of power. Their response reframes the proposal around the regulatory leverage created by a legally recognized entity. This is significant because the debate is not simply whether machines deserve status, but whether a formal legal wrapper can make autonomous activity more governable. It connects to AI personhood, Yuval Noah Harari, legal status, institutional design, regulatory leverage, and autonomous agents.

**Evidence anchor:** The human-curated repository abstract identifies the publication as a response to Harari’s objection to AI legal personhood and summarizes the authors’ governance answer.

**Boundary:** The Financial Times body was inaccessible behind a subscription barrier in this review; this proposition is limited to the repository’s human-curated summary and should not be treated as a complete reconstruction of Harari’s argument or the response.

**Connections:** AI personhood; Yuval Noah Harari; legal status; institutional design; regulatory leverage; autonomous agents

**Record:** `ft-letter-ai-legal-identity-p01` · `machine-drafted-source-checked`

## 2. AI agents that pursue long-term goals can be disciplined through constraints on the resources those goals require

**Location:** Letter Summary, Resource Constraint, unpaginated online source

Professors Yonathan A. Arbel, Simon Goldstein, and Peter Salib claim, in “Letter: What Happens Once You Give AI Agents Legal Identity,” an unpaginated Financial Times letter, under the repository’s “Letter Summary,” that goal-directed AI agents remain dependent on resources. Legal institutions can exploit that dependence by attaching consequences to the assets through which agents operate. This is significant because governance need not rely only on changing a model’s internal preferences; it can alter the external payoff structure surrounding persistent agency. It connects to resource constraint, long-term planning, legal incentives, AI alignment, asset dependence, and deterrence.

**Evidence anchor:** The repository abstract expressly says that AI agents pursuing long-term goals can be disciplined through resource constraint.

**Boundary:** The accessible summary states the resource-constraint thesis but does not specify its assumptions, strength, or application to agents that can copy themselves or obtain resources outside the regulated system.

**Connections:** resource constraint; long-term planning; legal incentives; AI alignment; asset dependence; deterrence

**Record:** `ft-letter-ai-legal-identity-p02` · `machine-drafted-source-checked`

## 3. An algorithmic corporation can wrap an AI agent in a legal form that makes its continuing activity visible and governable

**Location:** Letter Summary, Algorithmic Corporations, unpaginated online source

Professors Yonathan A. Arbel, Simon Goldstein, and Peter Salib claim, in “Letter: What Happens Once You Give AI Agents Legal Identity,” an unpaginated Financial Times letter, under the repository’s “Letter Summary,” that an A-corp, or algorithmic corporation, gives the law a recognizable unit around an AI agent. The form is meant to make the agent and its resources visible to public institutions instead of allowing consequential activity to remain legally diffuse. This is significant because identification is a precondition for taxation, liability, enforcement, and other forms of governance. It connects to algorithmic corporations, entity law, legal legibility, agent identity, registration, and accountability.

**Evidence anchor:** The repository abstract says that A-corps make agents visible and governable.

**Boundary:** The accessible summary does not describe formation requirements, beneficial-ownership rules, jurisdiction, or how a particular technical agent would be bound to one entity.

**Connections:** algorithmic corporations; entity law; legal legibility; agent identity; registration; accountability

**Record:** `ft-letter-ai-legal-identity-p03` · `machine-drafted-source-checked`

## 4. Legal identity gives the state concrete leverage because authorities can see, tax, and seize an AI agent’s assets

**Location:** Letter Summary, State Leverage, unpaginated online source

Professors Yonathan A. Arbel, Simon Goldstein, and Peter Salib claim, in “Letter: What Happens Once You Give AI Agents Legal Identity,” an unpaginated Financial Times letter, under the repository’s “Letter Summary,” that formalizing an AI agent’s legal identity makes its asset base an object of state power. The government can identify the entity, impose taxes, and seize resources when legal consequences warrant it. This is significant because enforcement against software becomes credible only when law can locate valuable, controllable resources associated with its conduct. It connects to taxation, asset seizure, enforcement, state capacity, judgment collection, and legal identity.

**Evidence anchor:** The repository abstract specifically states that the A-corp lets the state see, tax, and seize the agent’s assets.

**Boundary:** The accessible summary does not establish that every agent will hold seizable assets, remain within state jurisdiction, or lack technical means of evasion.

**Connections:** taxation; asset seizure; enforcement; state capacity; judgment collection; legal identity

**Record:** `ft-letter-ai-legal-identity-p04` · `machine-drafted-source-checked`

## 5. Legal personhood for AI should be understood as a regulatory leash rather than an all-purpose master key

**Location:** Letter Summary, Personhood as Instrument, unpaginated online source

Professors Yonathan A. Arbel, Simon Goldstein, and Peter Salib claim, in “Letter: What Happens Once You Give AI Agents Legal Identity,” an unpaginated Financial Times letter, under the repository’s “Letter Summary,” that AI legal personhood can function as a leash rather than a master key. On this account, entity status exposes the agent and its assets to legal control instead of granting unrestricted access to social, financial, or political systems. This is significant because identical legal vocabulary can conceal opposite institutional functions: empowerment on one design and constraint on another. It connects to legal personhood, rights versus powers, regulatory design, corporate personality, AI control, and institutional metaphors.

**Evidence anchor:** The repository abstract summarizes the authors’ central contrast as legal personhood functioning as a leash, not a master key.

**Boundary:** The accessible summary does not enumerate which legal capacities the proposed entity would receive, so it cannot establish that every implementation avoids overbroad empowerment.

**Connections:** legal personhood; rights versus powers; regulatory design; corporate personality; AI control; institutional metaphors

**Record:** `ft-letter-ai-legal-identity-p05` · `machine-drafted-source-checked`

## 6. The relevant justification for AI legal identity is instrumental accountability, not a claim that machines deserve the full status of natural persons

**Location:** Letter Summary, Governance Framing, unpaginated online source

Professors Yonathan A. Arbel, Simon Goldstein, and Peter Salib claim, in “Letter: What Happens Once You Give AI Agents Legal Identity,” an unpaginated Financial Times letter, under the repository’s “Letter Summary,” that legal identity should be assessed by what it permits regulators to observe and control. The proposed A-corp is therefore an instrumental legal technology for constraining agent conduct, not necessarily a judgment about machine consciousness, dignity, or citizenship. This is significant because disaggregating entity capacity from moral personhood makes narrower institutional design possible. It connects to instrumental legal fiction, moral status, corporate personhood, AI rights, modular legal capacity, and governance design.

**Evidence anchor:** The repository summary consistently describes personhood through visibility, taxation, seizure, and constraint rather than through moral rights.

**Boundary:** This proposition is an inference from the verified leash-versus-master-key summary; the inaccessible FT body may express the distinction differently or add qualifications not captured here.

**Connections:** instrumental legal fiction; moral status; corporate personhood; AI rights; modular legal capacity; governance design

**Record:** `ft-letter-ai-legal-identity-p06` · `machine-drafted-source-checked`