These closely related lawsuits are different, at least in degree, from previous lawsuits over allegedly defamatory AI output: Rather than alleging thoroughgoing hallucination, or (as in Battle v. Microsoft) the merging of two unrelated documents about similarly named people, they involve claims that AI output summarizing published documents overstates the allegations in those documents—a sort of claim that’s pretty common in normal libel litigation against newspapers.
1. The SEC charged Sergii Grybniak with securities with violating
[1] the antifraud provisions of Section 17(a) of the Securities Act of 1933 and Section 10(b) of the Securities and Exchange Act of 1934 and Rule 10b-5 thereunder, and
[2] the registration provisions of Sections 5(a) and (c) of the Securities Act ….
A federal court granted summary judgment to the SEC as to the registration violation allegations, and denied summary judgment to Grybniak as to the fraud allegations, on the grounds that “genuine disputes of material facts exist as to each category of [alleged] misrepresentations.” After that, the parties entered into a consent judgment. The SEC’s press release characterized the settlement thus:
In its complaint, the SEC alleged that … Grybniak and Opporty conducted an unregistered and fraudulent securities offering of crypto assets called OPP Tokens via an initial coin offering (“ICO”), raising approximately $600,000 from nearly 200 investors. The complaint also alleged Grybniak and Opporty marketed the ICO by making material misrepresentations and omissions and engaging in other deceptive conduct, including exaggerating the number of users and growth of its online blockchain-based marketplace for small businesses, misrepresenting the nature of Opporty’s purported partnership with a major software company, and claiming the ICO was “SEC regulated” and “100% SEC compliant” when it was not….
[T]he Court granted the Commission’s motion for partial summary judgment, finding Grybniak and Opporty had conducted an unregistered securities offering without a valid registration exemption in violation of Section 5 of the Securities Act. In addition, the Court rejected Grybniak and Opporty’s defenses asserting reliance on counsel and that they lacked fair notice of the application of the federal securities laws to the ICO.
Without admitting or denying the SEC’s allegations, Grybniak and Opporty consented to entry of the final judgment, which provides for permanent injunctive relief under Sections 5, 17(a)(2), and 17(a)(3) of the Securities Act. The final judgment also ordered Gryrbniak to pay a civil money penalty of $100,000, imposed a conduct-based injunction against him, and ordered both Grybniak and Opporty to comply with various undertakings.
As best I can tell, then, the SEC thought Grybniak committed fraud, but there was no court finding or admission as to that—though there was a court finding as to failure to register.
2. Grybniak then sued Google and X over AI outputs that, according to federal court rulings, Grybniak was “legally held responsible for making material misrepresentations and engaging in deceptive conduct” and that “according to … a federal court ruling, … Grybniak committed securities fraud and related violations.” Grybniak’s claim is that
Plaintiff has never been found liable for fraud, never admitted fraud, and has no criminal record. The single civil regulatory matter on which the statements are loosely based was resolved, on a no-admission basis, exclusively under non-fraud/negligence provisions of the federal securities laws—a registration provision and two negligence-based provisions that, as a matter of law, do not require and do not establish fraudulent intent.
The statements came in output from Google AI and X.AI (and allegedly continued to come after Grybniak informed Google and X about the errors in the output). Here’s a sample of the alleged output from Grok (the output from Google AI is different but similar):
Note that “Yes, according to the … SEC” seems quite true, but “… and a federal court ruling” seems to be incorrect.
The post Two New Large Libel Models Lawsuits, Though Alleging Mischaracterization Rather Than Outright Hallucination appeared first on Reason.com.
