Australian Mayor Threatening Lawsuit Over ChatGPT Libel

Financial Review (Australia) [Byron Kaye] reports (see also Reuters):

Brian Hood, who was elected mayor of Hepburn Shire, 120 kilometres north-west of Melbourne, last November, became concerned about his reputation when members of the public told him ChatGPT had falsely named him as a guilty party in a foreign bribery scandal involving a subsidiary of the Reserve Bank of Australia in the early 2000s.

Mr Hood did work for the subsidiary, Note Printing Australia, but was the person who notified authorities about payment of bribes to foreign officials to win currency printing contracts, and was never charged with a crime, lawyers representing him said.

The lawyers said they sent a letter of concern to ChatGPT owner OpenAI on March 21, which gave OpenAI 28 days to fix the errors about their client or face a possible defamation lawsuit….

Australian defamation damages payouts are generally capped around $400,000….

Comments

[Dr. Ed 2]

Can he sue a us company in an au court? If no physical presence in au of company?

[CDL]

Good question. The answer is: a defendant’s physical location (in the sense of locus of registered incorporation) is only one among a number of eligible criteria whose satisfaction can, alone or in combination, establish personal jurisdiction.

1. I’d assume the central jurisdictional criterion on which the plaintiff would rely in this matter is completion of the cause of action within an AU (NSW) jurisdiction. In the case of defamation over the internet, the High Court held in ‘Dow Jones & Company Inc v Gutnick’ (2002) 210 CLR 575 that, because the tort is complete when damage to reputation putatively arises, such damage manifests in the jurisdiction where the internet publication is published by being downloaded (or cached) on a computer (or device) situate in AU.

See: https://h2o.law.harvard.edu/text_blocks/5649

2. Service outside the jurisdiction is authorised because “damage” is claimed to be suffered wholly or partly in the jurisdiction.

3. Of course, the above is subject to a forum non conveniens “defence”. The general test in Australia is whether the AU forum is “clearly inappropriate”.

[CDL]

I was a bit sloppy in my annunciation of the ratio of Dow Jones v Gutnick, inasmuch as it applies the common law principle governing the identification of the jurisdictional locus of tortious wrongs. For clarification, see Gleeson CJ & McHugh, Gummow & Hayne JJ [2002] HCA 56 at paras [44]-[45]:

“[44] Reference to decisions such as Jackson v Spittall, Distillers Co (Biochemicals) Ltd v Thompson and Voth v Manildra Flour Mills Pty Ltd show that locating the place of commission of a tort is not always easy. Attempts to apply a single rule of location (such as a rule that intentional torts are committed where the tortfeasor acts, or that torts are committed in the place where the last event necessary to make the actor liable has taken place) have proved unsatisfactory if only because the rules pay insufficient regard to the different kinds of tortious claims that may be made. Especially is that so in cases of omission. In the end the question is “where in substance did this cause of action arise”? In cases, like trespass or negligence, where some quality of the defendant’s conduct is critical, it will usually be very important to look to where the defendant acted, not to where the consequences of the conduct were felt.

[45] In defamation, the same considerations that require rejection of locating the tort by reference only to the publisher’s conduct, lead to the conclusion that, ordinarily, defamation is to be located at the place where the damage to reputation occurs. Ordinarily that will be where the material which is alleged to be defamatory is available in comprehensible form assuming, of course, that the person defamed has in that place a reputation which is thereby damaged. It is only when the material is in comprehensible form that the damage to reputation is done and it is damage to reputation which is the principal focus of defamation, not any quality of the defendant’s conduct. In the case of material on the World Wide Web, it is not available in comprehensible form until downloaded on to the computer of a person who has used a web browser to pull the material from the web server. It is where that person downloads the material that the damage to reputation may be done. Ordinarily then, that will be the place where the tort of defamation is committed.”

[David Nieporent]

A more important question — well, not more important, I guess, but equally important — is whether an Australian plaintiff who wins a judgment against a U.S. defendant in an Australian court can collect on the judgment.

The SPEECH Act forbids U.S. courts from enforcing foreign libel judgments if the foreign country’s law doesn’t provide as much protection for speech as the 1A (or if the defendant’s acts were such that the defendant would’ve been liable under U.S. law.)

So if the AI company has no assets outside the U.S., this is probably nothing more than a symbolic gesture by the plaintiff.

[CDL]

Indeed. That US Act cannot but make critical the question of whether enforceability of the foreign judgment by a US court renders any court action commercially unreasonable. Perhaps there may be, however, some reasonable ground for anticipating that any judgment in favour of the AU plaintiff may be enforceable against the US defendant by a non-US court, depending on the jurisdictional connecting factors at play on the facts. I’ve no idea.

[CDL]

It doesn’t appear that the “Speech Act” (2010) – viz., 28 U.S. Code § 4102 – relevantly precludes a domestic US court from recognising and enforcing a foreign defamation judgment if (eg) the defendant was, in publishing the defamatory imputations, culpable of actual malice. My understanding is that 1A doesn’t supervene constitutionally to immunise speech against an action for defamation if the defendant knew the imputation was false, or was recklessly indifferent to its falsity. I may of course be wrong, given the complexity of the field.

Australia does recognise a Constitutionally-implied “freedom of political communication”, inferred from the entrenchment in the Commonwealth Constitution of a system of representative and responsible government. This implied freedom, however, is principally a constraint on Cth legislative & executive powers, not an individual right.

Still, it can operate derivatively in the context of private defamation (involving political communication) when any question of a defence of qualified privilege arises. Essentially, the High Court in Lange v ABC (1997) 189 CLR 520 accommodated the implied (negative) freedom to the common-law defence of qualified privilege by holding that, on a relevant occasion of reciprocal duty and interest, a publisher can be under a duty to publish the imputation if it be on political or governmental affairs (& thus qualifiedly immune from an action in defamation).

The problem the present ChatGPT case poses for the application of the “Lange” reformulation of qualified privilege (to accommodate “free speech”) is that the putatively defamatory imputations were “published” about a local government mayor. Doubtless they could generally be characterised as political, inasmuch as they could affect mayoral elections. But the Constitutional implied freedom operates specifically in respect of, & as a support for, the *Federal* system of representative & responsible government, & it’d be usually difficult to see how a local mayoral election, or local political affairs involving a sitting mayor, could (eg) have a practical effect on Federal elections, or on the affairs or political responsibilities of Federal ministers & officials.

And yet: this case is unusual inasmuch as the defamatory imputation – directly referring to a subsidiary of the RBA – does seem to have the real & not remote possibility of relating to a matter of high political moment at the Commonwealth level, involving the political/governmental duties & powers of responsible ministers & departments. The point – if the Lange “qualified privilege” defence can be applied – isn’t that the former RBA bribery scandal gives rise to any duty to publish the defamation because the political affair bears on the mayor & local politics, but that the existence of a *current* local mayor, in fact involved in bringing the RBA scandal to light albeit defamed by mere association with it, goes to the Commonwealth-level question whether the scandal had indeed been adequately addressed by Cth governmental officials.

Of course, even if the Lange-defence can apply, the privilege it supports is qualified & can be excluded by proof that the defendant (eg) was actuated by actual malice.

We thus arrive where we started: this case (if it is run in an AU court) may end up illustrating an exception to the “Speech Act” [?].

[Voize]

In the leading SPEECH Act case enforcement was denied because 1) under Canadian law truth is a defense, as in Australia, while in the US falsehood is an element of the tort that must be established by the claimant. That means the law is less protective, obviating the first exception. 2) regarding the merits criterion the US court doesn’t retry the case, it reviews the record, which did not support judgment under Mississippi law because the statement of claim was drafted to meet Canadian standards and so didn’t adequately plead falsehood.

US law excuses opinion defined broadly to include hyperbole and nonspecific conclusions as long as they don’t imply unstated facts. Australian “fair comment” is much narrower, often requiring explicit statement of the base facts and that the speaker honestly hold the opinion expressed. And of course there is the actual malice threshold the US applies to speech about any public figure.

The original intent of the SPEECH Act was to eliminate libel tourism but it applies to all foreign judgments, not just ones that should (or even could) have been litigated in the US. A US speaker on the Internet can do “defamation pirate radio”, libeling foreign targets knowing that they are protected from foreign judgments and that domestic courts won’t have jurisdiction if the plaintiff doesn’t have a US reputation to damage.

[CDL]

Thanks a lot for pointing out that case, & for your lucid epitome of it.

[Voize]

Australian defamation damages payouts are generally capped around $400,000

In Wilson v. Bauer Media (2017) the court found that the defendant’s actions justified exceeding the cap and plaintiff was awarded AUD$ 600K (reduced from $4.57M on appeal).

[A nerdy Fred]

I had wondered about this. OpenAI has been making their output available in places such as Australia and the UK where the libel laws are much more plaintiff-friendly than we’re used to.