Bundt Cake Recipes and Broken Locks
What one indictment reveals about federal document control
A federal grand jury in the Southern District of Florida indicted Carmen Mercedes Lineberger on Tuesday. She is 62. She just retired from the Department of Justice after nearly two decades, the last of them as Managing Assistant United States Attorney at the Fort Pierce branch office that prosecuted Donald Trump in the Mar-a-Lago classified documents case. She is alleged to have emailed Volume II of Jack Smith’s final report on that prosecution from her .gov account to her personal Gmail account on December 1, 2025. The file name she allegedly used was “Bundt_Cake_Recipe.pdf.”
Read that paragraph again.
Volume II is the special counsel’s narrative account of why he concluded the evidence would have sustained a conviction against Trump for the willful retention of national defense information. We aren’t allowed to read it. Judge Aileen Cannon entered an injunction the day after Trump’s second inauguration, enjoining the Department of Justice from releasing it outside the agency, and made the injunction permanent in February. American Oversight and the Knight First Amendment Institute have spent eighteen months trying to pry it loose through FOIA litigation, mandamus petitions, and intervention motions. The Eleventh Circuit told Cannon in November that she was unreasonably delaying. She delayed three more months and then ruled against release on a theory most legal scholars consider untenable.
That’s the document at the center of this case. “Bundt_Cake_Recipe.pdf.”
Here is the policy claim I want to make, plainly. The Lineberger indictment is the strongest argument for serious FOIA reform that anyone has handed the public in a decade. It is stronger than the Moynihan Commission report of 1997, which the political class has spent thirty years politely ignoring. It is stronger than the Public Interest Declassification Board’s 2020 Vision for the Digital Age. It is stronger than every think-tank white paper on government secrecy I have ever read, including the good ones. It is stronger because it is comic, and the public will remember a comic story more readily than a serious one. Stories about regulatory architecture do not travel. Stories about a federal prosecutor naming the most-litigated document in modern federal law enforcement “Bundt_Cake_Recipe.pdf” and emailing it to her Gmail travel account at the speed of a screenshot.
So I want to use the comic frame to make four serious points. Then I want to come back to the Bundt cake.
The system did not catch her.
That is the first point. The forensic detail every reform conversation should start from is that the alleged scheme is, technically, the most boring possible exfiltration. A PDF, renamed, attached to an outgoing email, sent from a .gov address to a Hotmail address in September and then to a Gmail address in December. Any halfway-competent enterprise data-loss prevention system can fingerprint a marked document and refuse to let it leave the network in an attachment. Any modern DLP can flag a federal prosecutor mailing PDFs from her work account to her personal account at all, let alone PDFs containing the contents of a document under a federal judge’s injunction. The Department of Justice either doesn’t deploy such systems across its U.S. Attorneys’ Offices, or it deploys systems so easily defeated that the trivial expedient of renaming a file to “Bundt_Cake_Recipe.pdf” is sufficient.
This is not a hypothetical. The conduct allegedly took place over ten weeks. Two separate transmissions. Hotmail in September, Gmail in December. According to the Department of Justice press release, the FBI’s investigation discovered it after the fact, post-retirement. The system did not catch her in real time. Lineberger is the named defendant, and if the allegations are true, she has earned her prosecution. But the institutional failure exposed isn’t principally a failure of one prosecutor’s integrity. It is a failure of the technical architecture that is supposed to make individual integrity unnecessary.
What “chocolate cake” tells you about “Bundt_Cake.”
The September email was sent under the subject line “chocolate cake recipe,” with the attachment named “Chocolate cake recipe.pdf.” The December email was sent under the subject line “Bundt_Cake_Recipe.pdf,” with the attachment named “Bundt_Cake_Recipe.pdf.”
The progression is the prosecution’s whole intent theory. In September, Lineberger sent internal DOJ communications and a memorandum marked “FOR OFFICIAL USE ONLY,” disguised as a chocolate cake recipe. The Department of Justice was not yet looking. Three months later, in December, she sent the sealed Volume II itself, disguised as a Bundt cake recipe. The escalation from casual lowercase (“chocolate cake recipe”) to formal underscored camel-case (“Bundt_Cake_Recipe.pdf”) tells you the strategy was rehearsed. It worked in September. So she ran it again in December, with greater operational confidence, and with the most contested judicially sealed document in the country.
This is what a federal prosecutor’s 18 U.S.C. § 1519 obstruction theory looks like in practice. The statute carries twenty years. Most of the press coverage of the Lineberger case has focused on the twenty-year number as if it were the punchline. It isn’t. The punchline is that the government chose § 1519 because the cadence of the file names, on the government’s theory, is evidence that Lineberger knew what she was doing and developed a strategy to keep doing it. The chocolate cake was a dry run. The Bundt cake was the real attempt.
There is a defense theory available, and Lineberger’s attorney has not yet articulated one publicly. The recipe filenames could be evidence of naïveté rather than concealment, a way to mark files for her own reference that only looks guilty in retrospect. She sent the files to her own personal accounts, made no attempt at encryption, and distributed them to no one outside. That is a real argument. But it is an argument that the trial process will have to engage with. The government’s reading is the more parsimonious one, and the file-naming pattern is the central evidence supporting it.
