The Volokh Conspiracy
Mostly law professors | Sometimes contrarian | Often libertarian | Always independent
Large Libel Models: An AI Company's Noting That Its Output "May [Be] Erroneous]" Doesn't Preclude Libel Liability
[An excerpt from my forthcoming article on "Large Libel Models? Liability for AI Outputs."]
AIs could, of course—and probably should—post disclaimers that stress the risk that their output will contain errors. Bard, for instance, includes under the prompt box, "Bard may display inaccurate or offensive information that doesn't represent Google's views." But such disclaimers don't immunize AI companies against potential libel liability.
To begin with, such disclaimers can't operate as contractual waivers of liability: Even if the AIs' users are seen as waiving their rights to sue based on erroneous information when they expressly or implicitly acknowledge the disclaimers, that can't waive the rights of the third parties who might be libeled.
Nor do the disclaimers keep the statements from being viewed as actionable false statements of fact. Defamation law has long treated false, potentially reputation-damaging assertions about people as actionable even when there's clearly some possibility that the assertions are false. No newspaper can immunize itself from libel lawsuits for a statement that "Our research reveals that John Smith is a child molester" by simply adding "though be warned that this might be inaccurate" (much less by putting a line on the front page, "Warning: We may sometimes publish inaccurate information"). Likewise, if I write "I may be misremembering, but I recall that Mary Johnson had been convicted of embezzlement," that could be libelous despite my "I may be misremembering" disclaimer.
This is reflected in many well-established libel doctrines. For instance, "when a person repeats a slanderous charge, even though identifying the source or indicating it is merely a rumor, this constitutes republication and has the same effect as the original publication of the slander." When speakers identify something as rumor, they are implicitly saying "this may be inaccurate"—but that doesn't get them off the hook.
Indeed, according to the Restatement (Second) of Torts, "the republisher of either a libel or a slander is subject to liability even though he expressly states that he does not believe the statement that he repeats to be true." It's even more clear that a disclaimer that the statement merely may be inaccurate can't prevent liability.
Likewise, say that you present both an accusation and the response to the accusation. By doing that, you're making clear that the accusation "may [be] inaccurate." Yet that doesn't stop you from being liable for repeating the accusation.
To be sure, there are some narrow and specific privileges that defamation law has developed to free people to repeat possibly erroneous content without risk of liability, in particular contexts where such repetition is seen as especially necessary. For instance, some courts recognize the "neutral reportage" privilege, which immunizes evenhanded reporting of allegations and responses in certain situations: "[W]hen a responsible, prominent organization … makes serious charges against a public figure, the First Amendment protects the accurate and disinterested reporting of those charges," even when the reporter has serious doubts about the accuracy of the charges. But other courts reject the privilege altogether. And even those that accept it apply it only to narrow situations: Reporting allegations and responses remains actionable—even though the report makes clear that the allegations may be mistaken—when the allegations relate to matters of private concern, or are made by people or entities who aren't "responsible" and "prominent." It certainly remains actionable when the allegations themselves are erroneously recalled or reported by the speaker.
The privilege is seen as needed precisely because of the general rule that—absent such a privilege—passing on allegations can be libelous even when it's made clear that the allegations may be erroneous. And the privilege is a narrow exception justified by the "fundamental principle" that, "when a responsible, prominent organization … makes serious charges against a public figure," the media must be able to engage in "accurate and disinterested reporting of those charges," because they are "newsworthy" just because "they were made."
Likewise, the narrow rumor privilege allows a person to repeat certain kinds of rumors to particular individuals to whom the person owes a special duty —such as friends and family members—if the rumors deal with conduct that may threaten those individuals. (This stems because from what is seen as the special legitimacy of people protecting friends' interests.) This is why, for instance, if Alan tells Betty that he had heard a rumor that Betty's employee Charlie was a thief, Alan is immune from liability. But the privilege exists precisely because, without it, passing along factual allegations to (say) a stranger or to the general public—even with an acknowledgement that they "may [be] inaccurate"—may be actionable.
Now a disclaimer that actually describes something as fiction, or as parody or as a hypothetical (both forms of fiction), may well be effective. Recall that, in libel cases, a "key inquiry is whether the challenged expression, however labeled by defendant, would reasonably appear to state or imply assertions of objective fact." A statement that obviously contains no factual assertion at all—as opposed to just mentioning a factual assertion about which the speaker expresses uncertainty, or even disbelief—isn't actionable. But neither ChatGPT nor Bard actually describe themselves as producing fiction, since that would be a poor business model for them. Rather, they tout their general reliability, and simply acknowledge the risk of error. That acknowledgment, as the cases discussed above show, doesn't preclude liability.
 Ringler Associates Inc. v. Maryland Casualty Co., 80 Cal. App. 4th 1165, 1180 (2000).
 Restatement (Second) of Torts § 578 cmt. e; see also Martin v. Wilson Pub. Co., 497 A.2d 322, 327 (R.I. 1985).
 Edwards v. National Audubon Soc'y, 556 F.2d 113 (2d Cir. 1977). A few later cases have extended this to certain charges on matters of public concern against private figures. Others have rejected the privilege as to statements about private figures, without opining on its availability as to public figures. See, e.g., Khawar v. Globe Int'l, Inc., 965 P.2d 696, 707 (Cal. 1998); Fogus v. Cap. Cities Media, Inc., 444 N.E.2d 1100, 1102 (App. Ct. Ill. 1982).
 Norton v. Glenn, 860 A.2d 48 (Pa. 2004); Dickey v. CBS, Inc., 583 F.2d 1221, 1225–26 (3d Cir.1978); McCall v. Courier-J. & Louisville Times, 623 S.W.2d 882 (Ky. 1981); Postill v. Booth Newspapers, Inc., 325 N.W.2d 511 (Mich. App. 1982); Hogan v. Herald Co., 84 A.D.2d 470, 446 (N.Y. App. Div. 1982).
 A few authorities have applied this privilege to accurate reporting of allegations on matters of public concern generally, but this appears to be a small minority rule. Barry v. Time, Inc., 584 F. Supp. 1110 (N.D. Cal. 1984); Tex. Civ. Code § 73.005.
 Edwards, 556 F.2d at 120. Likewise, the fair report privilege allows one to accurately repeat allegations that were made in government proceedings, because of the deeply rooted principle that the public must be able to know what was said in those proceedings, even when those statements damage reputation. But it too is sharply limited to accurate repetition of allegations originally made in government proceedings.
 Restatement (Second) of Torts § 602.
 Id. cmt. 2. Another classic illustration is a parent warning an adult child about a rumor that the child's prospective spouse or lover is untrustworthy. Id. cmt. 1.
 See, e.g., Martin v. Wilson Pub. Co., 497 A.2d 322, 327 (R.I. 1985).
 Takieh v. O'Meara, 497 P.3d 1000, 1006 (Ariz. Ct. App. 2021).
 See, e.g., Greene v. Paramount Pictures Corp., 813 F. App'x 728, 731–32 (2d Cir. 2020). Even then, a court might allow liability if it concludes that a reasonable person who knows plaintiff would understand that defendant's ostensible fiction is actually meant to be as roman à clef that conveys factual statements about plaintiff. The presence of a disclaimer wouldn't be dispositive then. See, e.g., Pierre v. Griffin, No. 20-CV-1173-PB, 2021 WL 4477764, *6 n.10 (D.N.H. Sept. 30, 2021).