On the morning of February 10, 2026, Bradley Heppner sat at the defense table in the Southern District of New York. On a seized laptop, federal agents had found 31 documents he had written himself using Anthropic’s Claude. He had typed his defense strategy into the tool. He had mapped his legal exposure. He had worked through the facts of a $150 million securities fraud investigation the way a careful defendant prepares for a consequential conversation with his lawyers. Then he had sent those documents to Quinn Emanuel Urquhart & Sullivan. He believed they were privileged. Judge Jed Rakoff disposed of both privilege claims from the bench in nine words: ‘I’m not seeing remotely any basis for any claim of attorney-client privilege.’ The written opinion followed eight days later, on February 17, 2026. United States v. Heppner, No. 25-cr-00503-JSR (S.D.N.Y. Feb. 17, 2026). The doctrine Rakoff applied was not new. Anthropic’s consumer terms disclosed data collection, model training rights, and the right to share user inputs with governmental regulatory authorities. Communications lacking confidentiality at creation cannot acquire it by transmission to counsel. The AI element was a fact. It was not a variable. What Rakoff’s opinion could not do, within the constraints of a single criminal case, was ask what the fullness of American legal thought would say about the structural conditions that produced Heppner’s loss. Not just whether the doctrine was correctly applied. Also: whether the doctrine is adequate for what technology has made possible, who bears the cost when it is not, and what no judge can fix from the bench. This document examines how the justices who shaped American law would have ruled if they had heard the argument. Every one of them would have affirmed Rakoff’s result. Two would have written partial dissents on the work product element only. What follows are the ten opinions most analytically relevant to the questions Heppner raises, supported by the controlling authority each justice would have invoked, followed by a complete summary of all 115 justices. #1 — Lewis F. Powell Jr. Associate Justice, 1972-1987 | Burger Court AFFIRM — Most Doctrinally Important Powell wrote the opinion Heppner’s defense invoked and is the only justice required to grapple with his own prior work. Controlling Authority: Upjohn Co. v. United States, 449 U.S. 383 (1981). Powell wrote the unanimous opinion extending privilege to corporate employees throughout the hierarchy, rejecting the narrow control group test. The animating principle: ‘An uncertain privilege, or one which purports to be certain but results in widely varying applications by the courts, is little better than no privilege at all.’ Id. at 393. The extension rested on two structural requirements — the employees owed duties to the corporation, those duties were enforceable — not merely on the functional utility of the communication to the representation. United States v. Kovel, 296 F.2d 918 (2d Cir. 1961). Judge Friendly extended privilege to accountants assisting attorneys, but only because attorneys engaged and directed those accountants. The third party owed confidentiality obligations through the attorney’s professional responsibility framework. Id. at 921-22. The Opinion Powell Would Write: ‘We do not retreat from Upjohn’s functional purpose analysis. We apply it. The defense argues that Claude served a functional purpose within the representation — the same way corporate employees in Upjohn provided information attorneys needed. That argument misreads Upjohn. The extension in Upjohn rested on the human relationship and the enforceable confidentiality duty, not on functional utility alone. Upjohn, 449 U.S. at 395. Claude has no duty. Anthropic has no confidentiality obligation to Heppner. The terms it published expressly disclaimed one. A tool that has contractually eliminated confidentiality cannot be analogized to a Kovel third party regardless of the functional similarity in the client’s mind. Kovel, 296 F.2d at 921. Heppner’s expectation was not reasonable. His privilege was not uncertain. It was absent.’ Upjohn, 449 U.S. at 393, applied. #2 — John G. Roberts Jr. Chief Justice, 2005-Present | Roberts Court AFFIRM — Most Consequential Opinion Roberts’s digital-age jurisprudence most directly frames what Heppner decides and — more importantly — what it does not. Controlling Authority: Riley v. California, 573 U.S. 373, 393, 396 (2014). Roberts wrote the unanimous opinion requiring a warrant before searching a cell phone incident to arrest. Cell phones hold ‘the privacies of life.’ Id. at 403. The physical fact of the arrest does not transform the digital content of the phone. The physical/digital distinction is structural, not circumstantial. Carpenter v. United States, 585 U.S. 296, 310 (2018). Roberts wrote the 5-4 opinion requiring a warrant for historical cell site location information. He acknowledged the third-party doctrine but held it required reconsideration for digital records: ‘The fact that the Government can compel a third party to disclose information does not mean the person sharing that information has no privacy interest in it.’ Id. The voluntary disclosure rationale operates with reduced force when third-party routing is a structural condition of communication rather than a deliberate choice. The Opinion Roberts Would Write: Roberts would affirm on the specific facts and write the most consequential separate analysis. Heppner voluntarily chose Claude. The government did not compel that choice. Heppner accepted non-confidentiality terms that expressly disclosed government access rights. Riley and Carpenter do not override voluntary disclosure of that kind. Heppner falls on the voluntary side of the line Carpenter drew. 585 U.S. at 310. ‘We note, however, that the question decided today is narrow. There is a distinction between a defendant who voluntarily accepts a ToS disclaiming confidentiality and an attorney who relies on enterprise-tier contractual protections while the government compels backend surveillance architecture that defeats those protections without notice to the client. Carpenter held that the third-party doctrine requires reconsideration when digital architecture makes third-party routing a structural condition of communication, not a deliberate choice. Id. Whether FISA Section 702 collection authority, exercised against platforms that have contractually promised confidentiality to enterprise clients, presents that question is not decided today. It is not outside the framework Riley and Carpenter established.’ #3 — Louis D. Brandeis Associate Justice, 1916-1939 | White, Taft, and Hughes Courts AFFIRM — Structural Warning Brandeis understood, better than any justice before or since, that legal fictions dressed in the language of voluntary choice conceal structural realities that doctrine cannot pretend away indefinitely. Controlling Authority: Olmstead v. United States, 277 U.S. 438, 471-78 (1928) (Brandeis, J., dissenting). The Court held 5-4 that wiretapping a private telephone did not constitute a Fourth Amendment search because no physical trespass occurred. Brandeis dissented alone: ‘They conferred, as against the Government, the right to be let alone — the most comprehensive of rights, and the right most valued by civilized men.’ Id. at 478. The Court adopted his position thirty-nine years later. Katz v. United States, 389 U.S. 347 (1967). Brandeis & Warren, ‘The Right to Privacy,’ 4 Harvard Law Review 193 (1890). The foundational text of American privacy law. Written sixteen years before Brandeis joined the Court. Brandeis, Other People’s Money and How the Bankers Use It (1914). Brandeis’s analysis of how financial intermediaries exploit asymmetric information. His framework: wherever power asymmetries allow one party to impose terms on another under conditions not genuinely voluntary, the law should examine what ‘voluntary’ actually means. The Opinion Brandeis Would Write: Brandeis would affirm on the narrow question without difficulty — Heppner accepted the terms; the terms controlled — and then write the opinion no other justice would write. Anthropic’s consumer terms reserve the right to disclose user inputs to ‘governmental regulatory authorities.’ That is not hypothetical. It is the architecture of the waiver. A legal system that calls government access to privileged communications ‘voluntary’ because the client accepted a standard-form contract he could not negotiate has redefined ‘voluntary’ in the way the author of Other People’s Money recognized immediately: when one party holds all the drafting power and the other party holds only the option to accept or forgo the service, formal consent is not the same as genuine choice. ‘This case involves a voluntary disclosure and the result follows from that fact. The case where the government compels the conditions making disclosure functionally inevitable remains to be decided. That case will come. When it does, the word voluntary will require examination this opinion cannot supply. The right to be let alone is the most comprehensive of rights. It does not become less comprehensive because the mechanism of its defeat is a ToS rather than a wiretap.’ Olmstead, 277 U.S. at 478 (dissenting). #4 — Robert H. Jackson Associate Justice, 1941-1954 | Stone and Vinson Courts AFFIRM — Most Analytically Durable Framework Jackson’s Youngstown concurrence is the most-cited framework for organizing ambiguous authority questions. It translates to AI privilege doctrine with a precision that suggests Jackson would have recognized the structural parallel immediately. Controlling Authority: Youngstown Sheet & Tube Co. v. Sawyer, 343 U.S. 579, 635-38 (1952) (Jackson, J., concurring). Jackson organized presidential power into three zones: Zone 1 — acting with express or implied congressional authorization, maximum authority; Zone 2 — acting without congressional guidance, the zone of twilight; Zone 3 — acting against congressional authorization, lowest ebb. Id. The framework organized ambiguous authority questions that the majority’s direct holding could not reach and became the governing structure for every subsequent separation-of-powers case. West Virginia Board of Education v. Barnette, 319 U.S. 624, 642 (1943). ‘If there is any fixed star in our constitutional constellation, it is that no official, high or petty, can prescribe what shall be orthodox.’ Id. Jackson’s model: establish structural principle first, apply it to specific facts second. Build the framework to outlast the case that generated it. Brown v. Allen, 344 U.S. 443, 540 (1953) (Jackson, J., concurring). ‘We are not final because we are infallible, but we are infallible only because we are final.’ Institutional authority requires structural consistency. Doctrine must be organized so courts can apply it reliably, not because courts are always right, but because consistency is itself a component of legitimacy. The Opinion Jackson Would Write: Jackson would affirm Heppner and write the Zone framework for AI privilege that every subsequent court would use. Zone 1: The attorney directs AI use within the representation, controls the architecture through enterprise-tier confidentiality protections, and the AI operates as the functional equivalent of a Kovel third party under attorney direction. Privilege holds. Upjohn, 449 U.S. at 395; Kovel, 296 F.2d at 922. Zone 2: The client uses AI without attorney direction, the terms contain no express non-confidentiality disclosure, and the communication could be characterized as preparatory to the representation. Zone of twilight. Fact-intensive. No bright line. Zone 3: The client uses AI on express non-confidentiality terms, without attorney direction, and the AI actively disclaims providing legal advice. Heppner. Zone 3. Heppner is Zone 3. He prepared 31 documents on a platform that told him his inputs were not confidential, without attorney direction, using a tool that disclaimed giving legal advice. Each element places him squarely in the zone where privilege is at its lowest ebb. ‘We affirm. We do so not because this case is difficult, but because the framework it requires is not. Zone 3 is Zone 3. The question courts will face is Zone 2. We leave them a structure.’ Youngstown, 343 U.S. at 635-38, applied. #5 — Potter Stewart Associate Justice, 1958-1981 | Warren and Burger Courts AFFIRM — His Own Standard Disposes of the Case Stewart wrote the test whose language most cleanly and directly disposes of Heppner’s privilege claim — and would have recognized the irony that the standard he created to protect individuals from government surveillance now applies against a defendant who chose the non-confidential instrument himself. Controlling Authority: Katz v. United States, 389 U.S. 347, 351-52 (1967). Stewart wrote the majority overruling Olmstead: ‘What a person knowingly exposes to the public, even in his own home or office, is not a subject of Fourth Amendment protection. But what he seeks to preserve as private, even in an area accessible to the public, may be constitutionally protected.’ Id. Harlan’s concurrence formalized the reasonable expectation of privacy test. Id. at 360-61 (Harlan, J., concurring). Berger v. New York, 388 U.S. 41 (1967). Stewart joined the majority striking down New York’s wiretapping statute, establishing procedural requirements for electronic surveillance that prefigured digital communications doctrine. The Opinion Stewart Would Write: Stewart would apply the Katz framework as doctrinal analogy, not constitutional holding — privilege analysis is common law, not Fourth Amendment. But the structure is identical. Heppner ‘knowingly exposed’ his communications to a commercial platform whose terms he accepted. He did not ‘seek to preserve’ those communications as private. He handed them to a service that told him, in terms he accepted, that they were not private. Katz, 389 U.S. at 351. He then transmitted them to counsel. The transmission added nothing. What lacked confidentiality at creation cannot acquire it by subsequent routing. Stewart would note the irony without softening it: Katz adopted Brandeis’s Olmstead dissent thirty-nine years after Brandeis wrote it. Brandeis dissented in Olmstead to protect individuals from government surveillance. The standard that dissent produced now governs when the individual chose the non-confidential instrument himself. Both results follow from the same principle. What matters is what the disclosing party did with the confidentiality of the commu

Originally published on LinkedIn Newsletter: The Technology Blind Spot