A Sony Doctrine to Protect AI LLMs and to Balance Copyright with Innovation

In an article published on Techdirt Ira P. Rothken wrote "The technological marvel of large language models (LLMs) like ChatGPT, developed by AI engineers and experts, has posed a unique challenge in the realm of copyright law. These advanced AI systems, which undergo extensive training on diverse datasets, including copyrighted material, and provide output highly dependent on user “prompts,” have raised questions about the bounds of fair use and the responsibilities of both the AI developers and users.

Building upon the Sony Doctrine, which protects dual use technologies with substantial non-infringing uses, I propose the TAO (“Training And Output”) Doctrine for AI LLMs like chatGPT, Claude, and Bard. This AI Doctrine recognizes that if a good faith AI LLM engine is trained using copyrighted works, where the (1) original work is not replicated but rather used to develop an understanding, and (2) the outputs generated are based on user prompts, the responsibility for any potential copyright infringement should lie with the user, not the AI system. This approach acknowledges the “dual-use nature” of AI technologies and emphasizes the crucial role of user intent and inputs such as prompts and URLs in determining the nature of the output and any downstream usage.

Understanding LLMs and Their Training Mechanism

LLMs operate by analyzing and synthesizing vast amounts of text data. Their ability to generate responses, write creatively, and even develop code stems from this training. However, unlike traditional methods of copying, LLMs like ChatGPT engage in a complex process of learning and generating new content based on patterns and structures learned from their training data. This process is akin to a person learning a language through various sources but then using that language independently to create new sentences. AI LLMs are important for the advancement of society as they are “idea engines” that allow for the efficient processing and sharing of ideas.

Copyright law does not protect facts, ideas, procedures, processes, systems, methods of operation, concepts, principles, or discoveries, even if they are expressed in copyrighted works. This principle implies that the syntactical, structural, and linguistic elements extracted during the training of LLMs fall outside the scope of copyright protection. The use of texts to train LLMs primarily involves analyzing these non-copyrightable elements to understand and statistically model language patterns....

Read the entire article on Techdirt.

Defendants in Fintech case ask the US Supreme Court to determine whether or not the constitutional right to confront witnesses in person in a criminal case is still alive in the metaverse era

On March 2nd, 2023 the defendants in a Fintech white collar criminal case, out of the Second Circuit, filed a Writ of Certiorari to the United States Supreme Court raising weighty consititutional issues arising out of the trial court permitting remote video testimony of a key prosecution corporate witness and defendants' inability to confront the witness in court. The Case is entitled Akhavan et al v. US.

Read the full brief to the United Supreme Court or the summary below.

CONSTITUTIONAL PROVISION INVOLVED

The Sixth Amendment to the United States Con- stitution provides, in relevant part: “In all criminal prosecutions, the accused shall enjoy the right ... to be confronted with the witnesses against him.”

...

In a federal white collar trial for bank fraud in the Southern District of New York, Petitioners contended that their system for processing credit-card transactions did not intention- ally misrepresent any information material to U.S. banks. See App., infra, 4a-6a. In attempting to prove both materiality and intent, the Government relied heavily on Visa’s transaction-approval policies. Id. at 54a. When prosecutors subpoenaed the company to testify at the March 2021 trial, Visa chose a 57-year- old corporate representative from California. Id. at 54a-55a. Citing travel-related health concerns com- mon to tens of millions of Americans, he moved to tes- tify by remote video teleconference rather than in person at trial. Id. The district court granted the re- quest under the Second Circuit’s permissive standard, noting that Petitioners would retain “almost all” the benefits of in-person confrontation. Id. at 64a.

Predictably, the remote testimony was marred by technical glitches, as when the Visa representative’s screen froze during cross-examination by Petitioners’ lead counsel. App., infra, 210a. Even when the technology was working, Petitioners were denied the opportunity to “[l]ook” the witness “in the eye” and make him assert allegations “to [their] face.” Coy, 487 U.S. at 1018-19. And the jury had no way to “draw its own conclusions” from mannerisms that cannot be observed remotely, such as a fidgeting foot or refusal to look at certain parts of the courtroom. Id. at 1019. Instead, this key prosecution witness—later alluded to 42 times in the Government’s closing argument— was a remote figure on a pixelated screen, appearing from the comfort of his attorney’s office 3,000 miles away from the crucible of the criminal trial.

...

The Sixth Amendment guarantees the accused “the right ... to be confronted with the witnesses against him.” This Court has long recognized that those words mean what they say: a criminal defendant must have the “opportunity to challenge his accuser in a face-to-face encounter in front of the trier of fact,” California v. Green, 399 U.S. 149, 156 (1970), subject to a sole, discrete exception applicable when “necessary” to enable testimony by a traumatized child sexual-abuse victim, Maryland v. Craig, 497 U.S. 836, 857 (1990). The Second Circuit, however, has invented a separate, sweeping exception applica- ble whenever a court finds that “exceptional circumstances” and the “interest of justice” weigh in favor of remote video testimony. United States v. Gigante, 166 F.3d 75, 81 (2d Cir. 1999). That is precisely the kind of “open-ended exception[] from the confrontation re- quirement” that this Court has repeatedly foreclosed. Hemphill v. New York, 142 S. Ct. 681, 690 (2022) (quoting Crawford v. Washington, 541 U.S. 36, 54 (2004)). Other courts of appeals have accordingly re- pudiated the Second Circuit’s position as an “outlier,” United States v. Carter, 907 F.3d 1199, 1208 n.4 (9th Cir. 2018), which “stands alone” in a lopsided circuit conflict, United States v. Yates, 438 F.3d 1307, 1313- 14 (11th Cir. 2006) (en banc).

The Court should grant review in this case to resolve that conflict. By allowing a corporate representative in a white-collar trial to testify remotely based on health concerns shared by tens of millions of Americans, the Second Circuit illustrated just how broad and malleable its Gigante exception is. The remote testimony in this case was critical; the Government invoked it repeatedly to prove the contested materiality and intent elements of the bank-fraud charges, and the Second Circuit did not suggest that an error in permitting the remote testimony would be harmless. Nor is this case a one-off. Absent this Court’s intervention, the Second Circuit may only slide further and further down the Gigante slope, per- mitting remote testimony in circumstances that no other federal court of appeals would. Even after the pandemic subsides, the vast discretion afforded by Gigante will persist, while the advances in videoconfer- ence technology inspired by COVID-19 will make remote testimony ever more enticing.

This Court’s intervention is thus urgently needed. “The simple truth is that confrontation through a video monitor is not the same as physical face-to-face confrontation.” Yates, 438 F.3d at 1315. Even the best video technology cannot replicate the intangible cues conveyed by in-person testimony, when a witness “stand[s] face to face with the jury in order that they may look at him, and judge by his demeanor upon the stand and the manner in which he gives his testimony whether he is worthy of belief.” Mattox v. United States, 156 U.S. 237, 242-43 (1895). Nor does testi- mony through a camera across the country fulfill the defendant’s basic right to be in the presence and look in the eyes of his accuser—a protection so fundamen- tal that it dates back to ancient times. Coy v. Iowa, 487 U.S. 1012, 1015-16 (1988). In short, while “[v]irtual confrontation might be sufficient to protect virtual constitutional rights,” it is not “sufficient to protect real ones.” Order of the Supreme Court, 207 F.R.D. 89, 94 (2002) (statement of Scalia, J.).

...

Other courts of appeals have squarely and ex- pressly rejected the Second Circuit’s position in Gi- gante. See Carter, 907 F.3d at 1208 n.4 (Ninth Circuit stating that it “agree[s] with the Eighth and Eleventh Circuits that Gigante is an outlier and that the proper test is Craig”) (citing cases).

The Eighth Circuit rejected Gigante’s approach in United States v. Bordeaux, 400 F.3d 548 (8th Cir. 2005). The defendant there argued that his confron- tation rights were violated when the district court “al- lowed the prosecuting witness to testify via closed- circuit television.” Id. at 552. Citing Gigante, the Government argued that the test in Craig did “not control” because Craig “involved a one-way closed-cir- cuit television system,” while its witness had testified via “a two-way system.” Id. at 553. The Eighth Cir- cuit rejected that argument, reasoning (as relevant here) that Craig governed. Id. at 554. The court ex- plained that “Gigante does not persuade us that ‘con- frontation’ through a two-way closed circuit television is constitutionally equivalent to a face-to-face confron- tation because it neglects the intangible but crucial differences between a face-to-face confrontation and a ‘confrontation’ that is electronically created by cam- eras, cables, and monitors.” Id. at 554-55. In partic- ular, the “virtual ‘confrontations’ offered by closed- circuit television systems fall short of the face-to-face standard because they do not provide the same truth- inducing effect.” Id. at 554.

The Eleventh Circuit sitting en banc likewise re- jected Gigante in Yates. 438 F.3d at 1313-14. The question in Yates was “whether witness testimony presented on a television monitor at a criminal trial

in Montgomery, Alabama, by live, two-way video con- ference with witnesses in Australia, violated the De- fendants’ Sixth Amendment right to confront the witnesses against them.” Id. at 1309. Reprising its and the Second Circuit’s position in Gigante, the Gov- ernment argued that the Eleventh Circuit “should not apply the Craig rule” because the testimony was ob- tained “by two-way video conference rather than one- way video conference.” Id. at 1312. The Eleventh Cir- cuit “reject[ed] this reasoning,” stated that “[t]he Gi- gante trial court should have applied Craig,” and added that “[t]he Second Circuit stands alone in its refusal to apply Craig.” Id. at 1313-14. The court ob- served that “[t]he simple truth is that confrontation through a video monitor is not the same as physical face-to-face confrontation,” that “the two are not con- stitutionally equivalent,” and that “[t]he Sixth Amendment’s guarantee of the right to confront one’s accuser is most certainly compromised when the con- frontation occurs through an electronic medium.” Id. at 1315 (citing Bordeaux, 400 F.3d at 554-55).

The Ninth Circuit similarly rejected Gigante in Carter. There, the court of appeals vacated convic- tions because a prosecution witness had been permit- ted to testify remotely “by two-way video, as she was seven months pregnant and unable to travel.” 907 F.3d at 1202. The Ninth Circuit “expressly h[e]ld” that “Craig’s two-part test applies to the use of two- way video testimony,” thereby “join[ing]” the “other circuits” that had addressed the issue and applied Craig. Id. at 1207-08 & n.4 (citing Yates and Bor- deaux, supra). The Ninth Circuit observed that, “[r]egardless of whether the video procedure is one- way or two-way, the defendant is being denied ‘a phys- ical, face-to-face confrontation at trial,’” id. at 1208 n.4

(quoting Craig, 497 U.S. at 850), and that “equating a two-way video procedure with face-to-face confronta- tion necessarily neglects” crucial “‘intangible ele- ments’ of confrontation,’” id. (quoting Gigante, 166 F.3d at 81).

The conflict between Gigante and the decisions of the Eighth, Ninth, and Eleventh Circuits are clearest because those courts reversed convictions upon reject- ing Gigante and instead applying Craig. But other federal courts of appeals and state appellate courts have also repudiated the Second Circuit’s position by concluding that Craig’s standard—not Rule 15—gov- erns analysis of a request for remote witness testi- mony by two-way video. See, e.g., United States v. Wandahsega, 924 F.3d 868, 879 (6th Cir. 2019); United States v. Abu Ali, 528 F.3d 210, 240-41 (4th Cir. 2008); Horn v. Quarterman, 508 F.3d 306, 319 (5th Cir. 2007); State v. Mercier, 479 P.3d 967, 976-78 (Mont. 2021); Lipsitz v. State, 442 P.3d 138, 140 (Nev. 2019); State v. Thomas, 376 P.3d 184, 195 (N.M. 2016); Bush v. State, 193 P.3d 203, 214-15 (Wyo. 2008).

The Second Circuit’s anomalous approach came into especially sharp relief during the pandemic. Dis- trict courts repeatedly cited Gigante to justify devia- tion from physical confrontation based on COVID-19 concerns (albeit in cases where the issue was not pre- served for appellate review). See, e.g., United States v. Avenatti, 2022 WL 103494 (S.D.N.Y. Jan. 11, 2022) (relying on health risk if the witness contracted COVID-19 to permit remote testimony despite the fact that the witness testified in person against defendant at other trials during the COVID-19 pandemic);

United States v. Donziger, 2020 WL 5152162 (S.D.N.Y. Aug. 31, 2020).

By contrast, jurisdictions that have split from the Second Circuit have regularly found that COVID-19 concerns do not justify exceptions to in-person con- frontation. See, e.g., United States v. Riego, 2022 WL 4182431, at *3-*4 (D.N.M. Sept. 13, 2022) (denying re- quest to testify remotely despite witness’s “chronic respiratory condition that leaves her more vulnerable to COVID-19”); United States v. Kail, 2021 WL 1164787, at *1 (N.D. Cal. Mar. 26, 2021) (denying request to testify remotely based on concerns arising from COVID-19 and the witness’s medical conditions).

...

This case affords a prime vehicle for this Court to resolve the question presented. Petitioners preserved their Confrontation Clause objections throughout the proceedings. App., infra, 32a. The district court is- sued a series of written opinions rejecting them. Id. at 32a-36a, 54a-64a. And the Second Circuit squarely resolved the Confrontation Clause question in its de- cision, which—while unpublished—treated estab- lished circuit precedent as dispositive. Id. at 10a-12a.

The Second Circuit’s willingness to allow remote testimony in this case—when the witness was a cor- porate representative who had basic travel-related health concerns widely shared by tens of millions of Americans—demonstrates just how broad and malle- able its “extraordinary circumstances” and “interest of justice” rule really is. App., infra, 10a-12a. It is equally clear that the courts of appeals that have re- jected Gigante would have applied the far more re- strictive “necessity” standard of Craig, 497 U.S. at 855. And there is every reason to believe that other courts would have reached a different result on these facts; indeed, district courts within sister circuits have routinely denied similar requests. See p. 25, supra. The case thus highlights the circuit conflict.

Finally, the decision to permit remote testimony by the Visa corporate representative here was critically important. The Second Circuit did not suggest that any error would be harmless, and the error here clearly was not. Cf. Coy, 487 U.S. at 1021-1022 (re- jecting harmlessness argument after finding Confron- tation Clause violation). The Government put Visa’s policies at the center of the trial by relying on them

for the materiality and intent elements of its novel, strained bank-fraud theory. See p. 6, supra. Espe- cially given that reliance, it was not too much to ask that Visa’s representative actually be at the trial— just like the other witnesses, the jurors, the judge, the lawyers, and the courtroom staff.

Instead, Elliott was allowed to testify from the comfort of his attorney’s office 3,000 miles away; to ex- ploit technological glitches and limitations to thwart effective cross-examination; to avoid the hostile glares of the defendants, the perceptive eyes of the jurors, and the solemnity of the courtroom environment; and ultimately to provide crucial evidence that led to the criminal convictions of both Petitioners for facilitating transactions that were lawful where they occurred and profitable for the banks that processed them. Al- lowing that testimony was a paradigmatic and pro- foundly prejudicial violation of the Confrontation Clause.

This Court’s review is warranted to afford due re- gard for a core, express constitutional guarantee and to ensure that criminal defendants standing trial in the Second Circuit are afforded no less rights than criminal defendants elsewhere in the country.

New ConsenSys Mesh NFT standard pays on-chain royalties to creators and collectors

Ira P. Rothken, attorney and TreeTrunk [EIP-4910] co-founder, also emphasized that "we made improvements in the NFT standard for artists' rights and collector protection from content verification to license management. It was important to improve on the way NFTs handle legal rights so that it’s no longer a game of guessing what the license agreement terms are, especially on secondary market transactions. Unlike the previous NFT standard, with TreeTrunk the license agreement is embedded in the NFT metadata itself. Using TreeTrunk, NFT collectors will be better informed on what rights they are getting, and artists can more easily control their intellectual property."

Read more here and here

Defendant files motion to dismiss an indictment in a case where the US claims that credit card merchant accounts on a cannabis platform violate the federal bank fraud statute

• The United States, in a novel case, is attempting to prosecute defendants for federal bank fraud arising out of obtaining merchant accounts to process Visa and MasterCard payments on a large cannabis e-commerce platform in California.
• Cannabis is legal under California state law. 
• The indictment alleges the cannabis merchant or named defendants chose wrong merchant category codes but under the Visa and MasterCard rules merchants don't pick merchant category codes, merchant banks have a non delegable duty under the credit card "network" agreements to make the choice, and there is no merchant category code for cannabis.
• The e-commerce platform at the center of the case is like an “Uber for cannabis” where items can be ordered for intrastate delivery.
• Forcing consumers to use cash instead of credit cards to purchase legal cannabis puts consumers at risk and is at odds with consumer protection in states where cannabis is legal.
• This public interest case can impact the use of credit cards to purchase cannabis in the United States.

On June 26, 2020 the defendant, Hamid Akhavan, filed a motion to dismiss the indictment in the case of US v. Akhavan in the Southern District of New York. Mr. Akhavan is represented in the case by Quinn Emanuel and Rothken Law Firm.

According to Ira Rothken, co-counsel for the defendant, “In our view, the government has brought a test case using federal law to try to criminalize the use of credit cards on cannabis e-commerce platforms. Cannabis is legal in states like California where the e-commerce platform operates intrastate. The federal government, in bringing such a test case, not only impinges on the sovereignty of California to legalize the intrastate distribution of cannabis but also on consumer protection by putting people at risk who may be forced to carry around a lot of cash instead of using the safety of credit cards to purchase legal cannabis.”

A summary of defendant’s motion dismiss is below (the complete motion is found here):

On March 9, 2020 a Grand Jury issued a one-count indictment (the “Indictment”) against Hamid Akhavan et al (together, the “Defendants”).  The Indictment alleges generically that “many United States banks are unwilling” to process transactions involving marijuana. (Indictment at ¶1); see also (Indictment at ¶12) (“[M]ost banks in the United States were unwilling to process credit and debit card transactions involving marijuana. . . .”).  It also alleges the Defendants and their California-based company provided a mobile platform to process marijuana-related transactions for California and Oregon purchasers through both debit and credit card payments.  (Indictment at ¶¶ 3–4).   

These payments were allegedly conducted through “payment networks” run by Credit Card Companies like Visa and MasterCard—entities not covered by the bank-fraud statute.  The Government asserts these companies “have rules that prohibit their credit cards from being used for marijuana purchases,” even though this is false in at least the cases of Visa and MasterCard. (Indictment at ¶5) (claiming these policies also apply to many debit cards).  The Government then explains violations of these alleged prohibitions may result in a merchant being terminated from the payment network.  (Indictment at ¶7-8).  Notably, the Government does not and cannot complain of any fraud against the Credit Card Companies.

The scheme outlined in the Indictment boils down to inducing, via allegedly incorrect merchant coding, some set of U.S. banks to process some transactions involving marijuana even while some U.S. banks would otherwise opt against doing so.

The only allegation in the Indictment that appears logically related to how the scheme might operate—although the Indictment does not spell out the connection—is the use of “merchant category codes” that are alleged to be deceptive.  (Indictment ¶9). 

However, merchants do not choose merchant category codes, the Visa and MasterCard merchant or "acquiring" banks do. Indeed, the merchant banks, under the Visa and MasterCard rules or network agreement have a non-delegable duty to pick the merchant category codes. The indictment does admit there was no merchant category code for cannabis. If "better" merchant category codes were selected within the Visa and MasterCard networks it would still not supply enough information to the credit card network issuing banks that any transaction involved cannabis. In other woads the merchant category codes cannot be used by issuing banks to filter or reject cannabis transactions. The bank fraud theory does not pass legal muster. The indictment simply assumes these codes are precise, well-understood warranties to the acquiring bank as to the exact subject matter of a particular transaction—such that use of an inaccurate or inapposite category code, or underlying sale of an item that has no applicable category code, translates to an intentional falsehood.  (Indictment ¶ 9).  But that blinks (or at least elides) reality.

The most notable aspect of the Indictment is what it does not allege.  First, there is no allegation of intent to cause loss or risk of loss, just as there is no allegation that any loss or risk of loss in fact resulted.  Second, there is no allegation of intent to illicitly obtain funds, as all transactions were in fact legitimate and intended by a consenting, fully informed buyer and seller.  Third, there is not even an allegation that any particular financial player to which the bank fraud statute applies would or could have known to refuse a more properly-coded transaction (using a more accurate code, to the extent one existed), refused a differently-coded transaction, or refused any transaction in the face of perfect information.  Here, the Indictment’s failure to include any specifics about the name or policy of any issuing bank (which are the only statutorily possible victims of the alleged scheme) draws into stark relief its impenetrable vagueness.

Inside the complex world of illegal sports streaming

In a Yahoo Sports article published today the author writes that on "[a] recent Sunday visit to one streaming site, whose owner told Yahoo Sports he transmits around 500 games per day, turned up functional feeds of second-division Guatemalan soccer, Polish women’s basketball, Canadian high school hockey and college baseball. A Tuesday visit to another site presented 49 different links to streams of a Champions League showdown between Liverpool and Bayern Munich.

The evolution is a product of many trends, from the progression of broadband to the growth of social media, to a generation of teens and young adults accustomed to enjoying content of all kinds for free. Together, they have birthed a burgeoning industry, propelled by an accelerating feedback loop. And they have enabled profit."

According to Ira Rothken "Illegal streaming is related to the options that one has to watch legal streams. In order to reduce the amount of illegal streaming, one has to have a society and culture that makes available legal streaming at a reasonable, affordable cost. … I suspect that content owners will provide better experiences, paid experiences, or free authorized experiences for users, and that will then lower the amount of piracy.”

The entire article can be found here.