The redacted version of Thomson Reuters’ reply to ROSS Intelligence’s list of Westlaw headnotes that it argued are not copyrightable is now available (see PDF below). Judge Bibas has already ruled that the jury should determine copyrightability of the headnotes, including their overall selection and arrangement in judicial decisions.
But Thomson Reuters’ reply is fascinating in two respects.
Examples of disputed headnotes
First, we get to see examples of headnotes that ROSS argued are not copyrightable, but that Thomson defended as copyrightable based on a minimal level of creativity. Let’s take a look at these examples. For ease of reading, I have included the text from the judicial opinions and the headnotes right below each example used in Thomson’s brief, with italics indicating the copied parts.
Example 1
Judicial opinion: To make a prima facie showing, the plaintiff “ordinarily cannot rest upon the pleadings, but is obliged to adduce evidence of specific facts.” Foster–Miller, Inc. v. Babcock & Wilcox Canada, 46 F.3d 138, 145 (1st Cir.1995); see Thomas v. Telegraph Publ’g Co., 151 N.H. 435, 437, 859 A.2d 1166 (2004). The trial court’s role, and our role in our de novo review, is “not as a factfinder, but as a data collector. That is to say, the court … must accept the plaintiff’s (properly documented) proffers as true for the purpose of determining the adequacy of the prima facie jurisdictional showing.”
Headnote: The trial court’s role when ruling, under the prima facie standard, on a motion to dismiss for lack of personal jurisdiction, and the appellate court’s role in its de novo review, is not as a factfinder, but as a data collector; thus, the court must accept the plaintiff’s properly documented proffers as true for the purpose of determining the adequacy of the prima facie jurisdictional showing.
Number of words copied: 45 words from 3 sentences of court’s opinion used by Thomson in 1 sentence. 21 words added by Thomson mainly to the beginning of the sentence.
Example 2
Judicial opinion: The inquiry should include whether the conduct calling for sanctions was willful and whether the other party was prejudiced by it, Vernell, supra; Brown v. Cohen, supra, 505 A.2d at 78–79; Garces, supra, 299 A.2d at 144–45; and the sanction imposed should, wherever possible, be tailored to the offense. Pollock v. Brown, supra, 395 A.2d at 52; see also Peek v. District of Columbia, 567 A.2d 50, 54 (D.C.1989); Vernell, supra (trial court abuses its discretion where the sanction is too strict or unnecessary under the circumstances).
Headnote: In determining whether to dismiss as sanction for failure to prosecute or to comply with superior court order, inquiry should include whether conduct calling for sanctions was willful and whether other party was prejudiced by it and sanction imposed should, wherever possible, be tailored to offense. Civil Rule 41(b).
Number of words copied: 28 words from 1 sentence of court’s opinion used by Thomson in 1 sentence. 21 words added by Thomson (18 not including the citation) mainly to the beginning of the sentence.
Example 3
Judicial opinion: By requiring the jury to find that Glen Mauldin accepted an item of value in return for official acts in counts five and nine, and that Tomblin promised something of value to Mauldin “with intent to influence an official act”, the instructions explain the reciprocity element of the quid pro quo. These instructions also distinguish between the levels of intent about which Tomblin is concerned, because they required the jury to find that, in promising something of value to Mauldin and in aiding and abetting Mauldin’s acceptance of an item of value, Tomblin had acted corruptly, that is, with unlawful purpose.
Headnote: By requiring jury to find that official accepted item of value in return for official‘s acts and that defendant promised something of value to official “with intent to influence an official act,” pattern instructions adequately explained reciprocity element of quid pro quo required under bribery statutes. 18 U.S.C.A. § 201; 5th Cir. Crim. PJI 2.12, 2.13.
[*I included the word “explain”/”explained” as the same, even though ROSS did not. However, unlike ROSS, I did not include the word “required” in the last part because the court’s opinion used that word for a different point than the prior sentence. Thomson objected to this kind of method, which it called a “Frankenstein” approach.]
Number of words copied: 37 words copied from 1 sentence from court’s opinion used by Thomson in 1 sentence. 8 words added by Thomson (excluding the citation).
Thomson vigorously defends the originality in its headnotes: “These West Headnotes are distillations, summaries, and/or otherwise reinterpretations of the judicial opinions that involve creative decision-making to create, which ROSS’s approach overlooks in painting them as verbatim copies. Such changes undoubtedly meet the low threshold for minimal creativity. See, e.g., CCC Info. Servs., Inc. v. Maclean Hunter Mkt. Reps., Inc., 44 F.3d 61, 6566 (2d Cir. 1994) (compendium of car valuation information was protectible); Eckes v. Card Prices Update, 736 F.2d 859, 863 (2d Cir. 1984) (decision as to which baseball cards to call premium was protectible); FMC Corp. v. Control Sol., Inc., 369 F. Supp. 2d 539 (E.D. Pa. 2005) (commercial product labels were protectible); West Publg Co. v. Mead Data Ctr., Inc., 799 F.2d 1219 (8th Cir. 1986) (arrangement of opinions in a case reporter was protectible).”
I must confess I didn’t find these 3 examples to be very good ones to show the putative originality in Thomson’s creation of individual headnotes, putting aside the selection and arrangement of them collectively in the judicial opinions. In the headnotes, there’s a high percentage of words copied straight from the judicial opinion. It’s true that Thomson added some more words (ranging from 8 words in the third example to 21 words in the first example and to 18 words in the second example), but, in each example, the majority of words in the headnotes are copied straight out of the judicial opinions. And, of course, the main points are derived from the court’s opinion, as you would expect to make them faithful to what the court actually held.
And I am not sure copying words from 3 sentences in a judicial opinion and sticking them into 1 sentence is creative. Nor do I think substituting “official” for the particular party’s name is creative. But, ultimately, the jury will decide. Possible it will find they satisfy the minimal level of creativity.
Did Thomson unintentionally bolster the case for recognizing authorship in ai Generated works with human contributions meeting “the low threshold for minimal creativity”?
Let’s assume Thomson is right that these 3 examples of headnotes satisfy the minimal level of creativity to be copyrightable. I think it may have consequences for how we think about the copyrightability of AI generated works. Thomson is advancing the position that their authorship of headnotes lies in distilling, summarizing, “or otherwise reinterpret[ing]” uncopyrightable judicial opinions and their activity “involve[s] creative decision-making to create.” It cites several decisions involving selection and arrangement and one involving product labels in support of the view that Thomson’s mental activity to create the headnotes “undoubtedly meet the low threshold for minimal creativity.”
Thomson stresses that this mental activity in creating headnotes, along with the overall selection and arrangement of headnotes, “requires judgment and choice.”
But, if that’s the case, I see no difference in that type of distillation and reinterpretation of uncopyrightable materials to create headnotes and the mental activity of humans required in prompt engineering AI generated works. If anything, prompt engineering is far less constrained, offering creators vast choices in deciding what to compose–a blank canvas. By contrast, condensing a judicial opinion into headnotes is constrained by what the court actually said. In any event, if writing headnotes is eligible for copyright based on distilling and reinterpreting uncopyrightable judicial opinions, a fortiori, prompt-engineered works should be, too.
Just as the underlying judicial opinion is uncopyrightable, the underlying AI generated work may be uncopyrightable (according to the Copyright Office’s position). But, in both cases, human creators may make creative choices that reinterpret what was first produced (by the judge or AI model, respectively), to produce something new and potentially original. To borrow Thomson’s words, as long as it “meet[s] the low threshold for minimal creativity,” the human contribution is copyrightable.
I have diagrammed below the comparison between creating headnotes and creating prompt-engineered works:
Uncopyrightable judicial opinion + human creative choices to reinterpret = headnote
Uncopyrightable AI work + human creative choices in prompts = AI generated work
For more of my views on AI authorship, I have a paper online.
To download Thomson Reuters’ brief on its headnotes, click below: