The Frontier: Person, Product, or Something New?
Why it Matters
A Florida corporation has the same powers as an individual to sue, be sued, and own land, and admiralty law has long treated a ship that commits an aggression as the offender. A monkey that took its own photograph could not sue under the Copyright Act, and Florida forbids its cities and counties to give a creek any legal rights, the right to flow included. In law, a person is whatever can hold rights and owe duties, and the law has made that call for reasons that have little to do with who can think or feel. Whether an AI system is a person, a product, or something new has run under this whole course, and so far the answer depends on which law is asking. You follow one open matter to its primary sources, teach it to the class, and say what the AI system in it should be called.
Current Context
On August 2, 2026, the European Commission gained the power to fine the makers of general-purpose AI models (opens in a new tab) under the AI Act, with fines that can reach 3 percent of a company’s worldwide turnover. On September 1 it sent its first requests for information to more than 30 AI providers (opens in a new tab), asking mostly about safety and copyright. On September 16, Microsoft AI’s chief executive, Mustafa Suleyman, answered the constitution Anthropic uses to train its Claude models (opens in a new tab), which treats a model’s moral status as an open question. AI systems are not conscious, he wrote, and controlling a model trained to believe it may be conscious might prove impossible. The law in Europe now treats these models as things their makers answer for, while the makers disagree in public about whether a model could ever be more than that.
Key Concepts
Reading the Record
Every legal development leaves a record where it happened: a case’s docket, with each filing and order numbered and dated; a bill’s text and history; or a rule’s file of notices, comments, and committee reports. The record shows where a matter stands on the day you read it. A story about the matter shows where it stood when the reporter wrote.
A Development and a Headline
A development is a dated event you can find in a primary source: an order entered, a bill referred to committee, a rule published, a fine imposed. A headline is a story about one, a secondary source that can lead you to the development and cannot stand in for reading it. When the two disagree, the primary source controls.
Legal Personhood
The status of being able to hold legal rights and owe legal duties: to own property, make contracts, sue and be sued, and answer for harm. The law grants it for its own reasons, which is why a Florida corporation has “the same powers as an individual” and a creek in Orange County has none. It is a legal status, not a finding that something can think or feel.
Product or Tool
The two categories most law now uses for an AI system. As a product, a system can be defective, and its maker can answer for the defect; as a tool, what it does counts as the act of the person using it. Florida’s electronic transactions law takes the tool view: a computer program it calls an electronic agent can form a contract that no person reviewed.
Electronic Personhood
A proposed third category, a limited legal status for the most autonomous machines so that the machine itself could answer for damage it causes. In 2017 the European Parliament asked the European Commission to consider it, alongside compulsory insurance and a compensation fund, and the AI Act the EU adopted in 2024 does not create it. Idaho’s code forbids granting personhood to AI, and an Ohio bill would do the same.
The Pacing Problem
The gap between how fast a technology changes and how fast law can answer it. Courts decide the case in front of them, legislatures meet on a calendar, and agencies take comments before they act, so law usually meets a new technology with categories built for older ones. The gap is why most matters on a watch list of AI law are still pending.
Four Modalities of Regulation
Lawrence Lessig’s name for the four constraints that shape what anyone does: law, social norms, the market, and architecture, which in software he calls code. A change in one moves the others, so a rule no legislature has written may already be enforced by a price, a custom, or a system’s design. In his terms, a developer that trains its model on a written set of values regulates the model by architecture.
Explaining to a General Audience
Stating a legal change so that someone outside the field understands what happened, why it matters, and what comes next, without losing accuracy. It takes plain words and no fact the speaker has not checked. Paralegals do it for clients every week.
Resources
- The Line: AI and the Future of Personhood, James Boyle, Faculty Author Celebration, Duke University School of Law (Oct. 2024) (opens in a new tab) (The video under the title, from 29:34 to 34:08. A Duke law professor explains why the law made corporations persons and how the same reasons could apply to AI. The European draft he mentions was the European Parliament’s 2017 proposal of “electronic persons,” which the EU never adopted. At 33:17 he says a court reporter made up the 1886 statement that corporations are persons under the Fourteenth Amendment. The report records it as the Chief Justice’s remark before argument, and the Court’s opinion did not decide the point.)
- Code: Version 2.0, Lawrence Lessig (2006) (opens in a new tab) (The author’s page, with a free download under a Creative Commons license. Read chapter 7, “What Things Regulate,” pages 120 to 125, where law, norms, the market, and architecture each act on the same choice.)
- Thaler v. Perlmutter, 130 F.4th 1039 (D.C. Cir. 2025) (opens in a new tab) (The one point the courts have settled: the Copyright Act requires a human author, and it treats machines as tools that authors use. The Supreme Court declined to hear the case on March 2, 2026.)
- Wilde Cypress Branch v. Hamilton, No. 6D23-1412 (Fla. 6th DCA Jan. 19, 2024) (opens in a new tab) (Three pages. Orange County’s charter gave its lakes and streams a right to flow, and the court held that section 403.412(9)(a), Florida Statutes, which bars local governments from granting legal rights to nature, preempted it.)
- Section 668.50, Florida Statutes, Uniform Electronic Transaction Act (opens in a new tab) (Florida’s definition of an electronic agent, in subsection (2)(f), and its rule in subsection (14) that electronic agents can form a contract no person reviewed.)
- House Bill 469, 136th General Assembly, Ohio House of Representatives (opens in a new tab) (A bill that would declare AI systems nonsentient; bar them from personhood, marriage, property, and corporate office; and put liability for their harms on owners, users, and makers. The page shows the bill’s status and links its text.)
- Claude’s Constitution, Anthropic (Jan. 22, 2026) (opens in a new tab) (One developer’s account of the values it trains into its models. Its section on Claude’s nature calls such systems “a genuinely new kind of entity” and leaves their moral status open. It states a company’s position, not the law.)
- RECAP Archive, CourtListener (opens in a new tab) (Free copies of federal dockets and filings. Search by case name or docket number for a federal item on the watch list.)
- Appellate Case Information System, Florida courts (opens in a new tab) (Florida’s appellate dockets, free and with no login. Choose the court, then enter the case number.)
- Bill List, The Florida Senate (opens in a new tab) (Florida bills with their text and history, including those already filed for the 2027 session.)
- Congress.gov (opens in a new tab) (Federal bills with their text, every action taken on them, and their committee reports, for an item such as the NO FAKES Act.)
What to Do
Most of what this course has covered is still moving, and a paralegal is often the person asked where a matter stands today. You take one item from the course’s watch list and find out from its own record what has happened since the list was printed. Then you teach it to the class in a short brief and a five-minute recording. The brief ends by saying what the law should call the AI system in your item, and why.
Part 1: Read and Choose
Start with the background reading. Then watch the video under the title from 29:34 to 34:08, where a Duke law professor explains why the law made corporations persons and how the same reasons could reach AI. Read pages 120 to 125 of chapter 7 of Lessig’s Code, where four kinds of constraint act on one person’s choice. Then choose one item from the watch list below. The list was printed on October 1, 2026, so within a day of choosing, find out whether your item has moved: a ruling issued, a hearing held or reset, a bill filed or dead. Write down the last dated event you can confirm in a primary source, and where you confirmed it.
The watch list, October 1, 2026. Dated items first, then matters waiting on a ruling or a vote with no date set.
October 13, 2026. Perplexity’s answer or motion to dismiss is due on Amazon’s amended complaint over an AI shopping agent, in Amazon.com Services LLC v. Perplexity AI, Inc., No. 3:25-cv-09514 (N.D. Cal.).
October 13, 2026. The last day for Ross to petition for rehearing of the Third Circuit’s September 29 decision that copying Westlaw’s headnotes to build an AI research tool was not fair use, in Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc., No. 25-2153.
October 15, 2026. The federal Advisory Committee on Evidence Rules meets to decide whether to republish proposed Rule 707, on evidence produced by AI, and whether a rule for deepfakes is needed.
October 26, 2026. Colorado’s Attorney General holds a hearing on the rules for Colorado’s AI law, SB 26-189, which applies to decisions made on or after January 1, 2027.
November 2026. The EEOC plans to rescind the Uniform Guidelines on Employee Selection Procedures, 29 C.F.R. part 1607, the source of the four-fifths rule.
November 15, 2026. The first payments to authors are due under the $1.5 billion settlement in Bartz v. Anthropic PBC, No. 4:24-cv-05417 (N.D. Cal.).
November 20, 2026. The reply briefs, the last summary-judgment briefs, are due in In re OpenAI, Inc. Copyright Infringement Litigation, No. 1:25-md-03143 (S.D.N.Y.), on whether training on news articles and books is fair use.
December 2026. Ohio’s 136th General Assembly ends, and HB 469, which would declare AI systems nonsentient, dies if it has not passed.
December 17, 2026. The court hears OpenAI’s motions, due October 15, to dismiss the Tumbler Ridge suits for forum non conveniens, the doctrine that lets a court send a case to a better-suited forum abroad. The lead case of 38 is Stacey v. Altman, No. 3:26-cv-03701 (N.D. Cal.).
March 2027. Florida’s regular legislative session opens. Watch whether SB 482, the 2026 Artificial Intelligence Bill of Rights, returns, and whether the Attorney General’s proposal to make companies criminally liable when their AI systems take part in crimes becomes a bill.
March 9, 2027. The court hears the motion for class certification in Mobley v. Workday, Inc., No. 3:23-cv-00770 (N.D. Cal.), on AI screening of job applicants.
No date set. A ruling on OpenAI’s motion to dismiss in Nippon Life Insurance Co. of America v. OpenAI Foundation, No. 1:26-cv-02448 (N.D. Ill.), on whether ChatGPT practiced law.
No date set. A Senate vote on the NO FAKES Act of 2026, S. 4591, which would create a federal right in a person’s voice and likeness.
No date set. The defendants’ response after remand, and a ruling on the Attorney General’s September 28 motion for a temporary injunction, in Florida’s suit against OpenAI, Office of the Attorney General v. OpenAI Global, LLC, No. 26-000295-GC-AXMX (Fla. 10th Cir. Ct.).
No date set. A ruling on OpenAI’s motion to dismiss, filed September 25, in the first of six product-liability suits in the federal court in Tallahassee, No. 4:26-cv-00222 (N.D. Fla.).
No date set. A ruling on qualified immunity in Dillon v. City of Jacksonville Beach, No. 3:26-cv-01930 (M.D. Fla.), over an arrest that followed a face-search match.
No date set. The Federal Trade Commission’s first enforcement action under the removal duty in section 3 of the TAKE IT DOWN Act.
Part 2: Build the Record
Use the Source Finder to learn where your item’s primary sources live. Then open and read them yourself: the docket and its latest order, the bill’s text and its last action, or the committee’s agenda book or report. Federal dockets are free on the RECAP Archive, Florida appellate dockets on the Appellate Case Information System, and federal and Florida bills on Congress.gov and the Florida Senate’s site. Build a source list with one line per source: what it is, its date, and the fact it establishes. A news story or tracker may lead you to a source, and it may appear on the list as how you found it. A fact that appears only there stays out of your brief.
Part 3: Write the Brief
Write a brief of 400 to 500 words for a classmate who has not followed your item. It has four parts. The first says what happened and when, and the second why it matters to someone who works in law. The third says what the primary sources say, with a page, section, or docket entry for each point, and the fourth what to watch next, with a date if one is set. Close with a paragraph on the question in this page’s title. Say what the law in your item treats the AI system as: a person, a product, a tool whose acts belong to its user, or something the law has not yet named. Then say whether that category fits and why, with at least one reason drawn from your sources.
When the brief is done, score it yourself on six lines worth up to 2 points each. The first four lines are its four parts. Score each 2 if every fact in the part traces to a source on your list, 1 if one fact does not, and 0 if more do not or the part is missing. The fifth is the source list, scored 2 if every source has its date and the fact it establishes, 1 if one does not, and 0 if more do not. The sixth is the closing paragraph. Score it 2 if it names the category the law now uses and argues for one with a reason from your sources, 1 if it does only one, and 0 if neither.
Part 4: Check the Brief
Run the Fact Verifier on each sentence in the brief that states a date, a number, a vote, or what a court, legislature, or agency did, pasting the sentence with the source you relied on. Open the source yourself before you fix or cut anything it marks. Then paste your closing paragraph and your source list into the Position Tester and answer its questions. Revise the paragraph only where an answer showed you a weakness. Under the brief, list what you changed after each check, and the one question from the Position Tester you found hardest to answer.
Part 5: Record the Explainer
Record an explainer of about five minutes for the same classmate. Script it yourself from the brief, so that every fact in the recording is in the brief and nothing new enters because it sounds good aloud. Say what happened, why it matters, and what to watch next in plain words, and end with your answer to the question in the title. Turn on captions if your recording tool offers them.
Part 6: Post and Reply
Post the brief with its scores, the source list, your notes from Part 4, and a link to the recording, all as a single new discussion thread. Then make your two replies, as the Posting and Reply Guidelines describe, to classmates who chose items other than yours. In each, add one fact about their item from a primary source their list does not include, with its citation and date. Run each reply through the Fact Verifier before you post it.
Part 7: Share Chat Link
Include one AI chat link with a 1–2 sentence explanation of what the conversation shows and why you chose to share it.
Suggested AI Prompts
Use these prompts as a starting point, then adjust them to fit your goal. Strong prompting develops through trial, revision, and testing. It’s a foundational skill that grows into more advanced AI work such as context engineering and agent-based workflows.
Source Finder
Act as a law librarian. I will name a pending case, proposed rule, bill, or regulatory action involving artificial intelligence. Tell me where its primary sources live and how to reach them: the court and docket number and whether the docket is on CourtListener or PACER, the bill number and its page on Congress.gov, the rulemaking committee and the uscourts.gov page for its agenda books and reports, or the Federal Register citation. Give me search terms rather than guessing at document numbers or dates, and tell me clearly when you are not sure a document exists or when your information may be out of date. Do not summarize the development; I will read the sources myself.
Role assignment with a narrow job makes this prompt useful: the model points you to the record and never tells you what is in it. Confining it to locating sources and admitting uncertainty about dates keeps its stale knowledge from becoming your brief’s claims.
Fact Verifier
Act as a fact-checker for a legal newsletter. I will paste a short claim I intend to post about a pending case, rule, or bill, along with the primary source I am relying on. Tell me whether the source supports the claim exactly as written, quoting the language that does, and flag any word in my claim that goes beyond the source, such as a date, a holding, or a prediction. If the source does not support the claim, say so. Use only the source I provide.
Source grounding limits the model to the text you paste, so it can tell you what the source supports but cannot supply a fact of its own. Pairing every claim with its source is the discipline the whole course has taught, applied to the sentences you are about to publish.
Position Tester
Act as a law professor who teaches a seminar on legal personhood. I will paste the closing paragraph of my brief on a pending AI matter, where I say whether the law should treat the AI system in it as a person, a product, a tool whose acts belong to its user, or something new, followed by the list of primary sources my brief relies on. Do not tell me which answer is right, and do not rewrite my paragraph. Ask me one question at a time, no more than five in all, and wait for my answer before asking the next. Each question should press my position with the strongest case for a category I did not choose, drawing on how the law has treated corporations, ships, animals, and rivers, on product liability, and on Florida’s rule that a computer program can act as a person’s electronic agent in forming a contract. After my last answer, tell me which of my answers was weakest and why, and name any statement about the law in my paragraph that I should check against a primary source before I post it.
Socratic elicitation keeps the argument yours, because the model asks and you answer, so the reasoning in your final paragraph is reasoning you had to defend. Holding it to one question at a time, each from a category you rejected, stops it from lecturing and aims it at the weak side of your position.