IPSC Breakout 5 Comparative Approaches

Stefania Fusco (co-author Valerio Sterzi), Does the EU Need an EBay-Like Case? Evidence Against Granting Automatic Injunctions in Europe

Patent paper; issues involving NPEs similar to trolling issues in US. If most infringement cases are between noncompetitors, automatic injunction is inappropriate and an eBay-like rule is appropriate, and their research suggests that this is the case.

Justin Hughes, Comparative Online Bad Guys

After issues of application of © online were settled, key was determining liabilities of intermediaries. A problem of elephants and mice. Enforcing against mice is difficult, but elephants can’t hide and are subject to lawsuits. So enforcement for CSAM through © went to platforms. DSM Directive in EU and US developments were all about enforcement on the elephants. A big part has been site blocking: a court or admin body orders an ISP to block internet access to particular online locations. Roots as old as 2001 Directive ordering methods developed, as well as 512(j)(1) allowing site-blocking injunctions in the US which has never really been tested in court. Now picking up speed—over 30 jurisdictions, most democracies, have endorsed site-blocking, often by the highest court. [defined: no fault injunctions against ISPs, so I guess China’s Great Wall doesn’t count?]

Early days: there were questions of power to make orders against innocent third parties. EU: statute. Canada: determined as matter of common law. EU got proportionality analysis. Early cases, also questions of technical possibility and burden sharing for site blocking. Can be done with domain names, IP addresses, a combination, URL blocking. 2014: ECJ said that the court didn’t need to designate technical measures—could order an outcome prohibition; transmission entities have to effectively achieve site-blocking. EU study: most EU jurisdictions have included that intermediaries must pay, also true in Australia. 2018: UK House of Lords, TM site-blocking case disagreed.

How do you ID the bad guys? Pirates, rogue, flagrantly infringing online locations (Singapore), infringement-based business models. This wasn’t an issue in early cases because if you’re asking to block “the pirate bay” there’s not a big problem, and courts will gloss over the necessary evidence. But more important with dynamic injunctions, where jurisdictions adopt some shortcut mechanism whereby an initial injunction against ISPs can be used by Ps to add URLs/domain names/IP addresses to the injunction. Adopted in Argentina, Canada, India, Singapore, UK, Netherlands, Germany.

Comparatively, jurisdictions are starting to develop criteria: Singapore’s “flagrantly infringing online location,” codified 10 years ago. Primary purpose to infringe ©; directories, indexes, categories of means to commit or facilitate © infringement. Australia adopted a similar law (includes weighing of harm to others and whether other injunctions have been issued). Good to look at what people in other countries are doing. Delhi high court decision 2019 followed Singapore/Australian list. Canada has considered use of new domain names, IP addresses, etc. to evade site blocking; India and Singapore consider volume of traffic/frequency of access to the online location; etc.

Many of the things the courts say are separate factors are going to “is the primary purpose or effect to cause © infringement?” Directories/indexes for promoting infringement; flagrancy; silence or inaction on receipt of takedown notices; circumvention instructions. One interesting factor: looking to whether website has been disabled by orders from one or more courts in other jurisdictions on © grounds. Australia, Denmark, India, Singapore all consider this.

Would not be surprised to see legislation coming to US this year. Hope will consider these models before adopting rules.

Q: how related to SOPA/PIPA?

A: motivated grassroots action—truly believe that American court could already do this under 512(j). Why didn’t people test 512(j) in court first? The answer is: most people in DC think that the tech community is ready for this. We’re in a very different place: now Google can do no right on Capitol Hill. Balance of power has changed.

Marketa Trimble: Under Nevada law, have to comply with other countries’ laws to be licensed for online gambling in Nevada.

A: hard to imagine US court doing this on its own initiative.

Peter Yu: duration of blocking? UK careful about that.

A: interesting but not important b/c pirates move on so quickly that an 18 month period is 5 business models. Might matter to transmission ISP to clean up their blocklists though.

Cesar Ramirez-Montes, Digital User Rights in the Mexican Supreme Court

Potential for global South to think of exceptions/limitations as user rights, using Mexico as a point of reference. Not bashing the country and its © system as such, but presenting the academic argument for room in global South to develop this notion, and Mexico specifically. Implementation of safe harbor provisions under USMCA was challenged before Mexican SCt; two different rulings.

Descriptively: USMCA provides only for notice and takedown, lifted from DMCA. Prescriptive: Mexican legislators should rethink their adoption of DMCA Plus mechanisms such as notice and stay down which was adopted as part of transposing the USMCA into legislation. Also: Mexican SC should develop notion of exceptions and limitations as user’s rights as a new rule of interpretation that provides greater attention to user interests and thereby balance © scope with larger public interest.

Almost excessive attention paid to rightholders’ interests and little interest in users’, audience’s, and public’s interest. As if it was merely a system of protecting authors and nothing else.

The public bears a great cost from this, and © owners gave nothing in return. Model: Canadian SCt ruling and relevant scholarly explorations of exceptions and limitations as user rights.

Mexico adopted the safe harbor framework from USMCA, which was modeled on DMCA. But added staydown obligation: “reasonable measures” to prevent same content from being uploaded. Implementation also covers TPMs though overrides protections for out of copyright works. Also funneling WIPO internet treaties, Marrakesh Treaty, and TPPA implementation. Result: DMCA-Plus approach to safe harbors.

Challenges before SCt: (1) freedom of expression under Mexican constitution, contrary to due process and legal principle; (2) lack of definition of key terms did not provide legal certainty. Notification requirements didn’t require evidence of infringement; chilling effects and failed to provide adequate remedies for users entitled to circumvent b/c of tool prohibitions on DRMs: a disabled person could be entitled to an exception but unable to get a device that would do the circumvention.

Jan. 2024, Mex. SCt First Chamber dismissed the appeal re TPMs but found that staydown unduly restricted freedom of expression according to proportionality test. Previously, Mexican law already had remedies for © owners. Comparative analysis of US law; cited EFF on DMCA’s effects. Grand Chamber arrived at a different conclusion in May/June 2024, but have not published reasoning.

Marketa Trimble, Cross-Border Exceptions and Limitations to Copyright: “Powered by AI”

Not interested here in whether training infringes/using AI as an example of a process. Legislation of exceptions and limitations is often considered on a single-country basis. But in multi-country transactions, such as when a library in country A wants to provide an article to a researcher via a library in country B, we need to know if there are identical or similar exceptions in A and B that cover the acts in the same manner. Libraries need cross-border exceptions; the model of AI suggests that we could get them, but libraries lack the political clout of big companies. Potential solutions: internationally uniform exception; mutual recognition of national exceptions (exists in some countries—once it’s permitted in one country, the other country accepts it as legal); mutual recognition with national verification (we have the same kind of exception and we check for compliance with our exception). Localization (no matter what happens across the border, we apply that country’s exception). Private ordering. Delocalization by brute force: Seen online in other areas where companies don’t bother to localize TOS, just comply with US law which is home country law; that solves the problem for them and they wait for litigation if it materializes. Technological solutions?

International law and EU law are models. Mutual recognition in EU Orphan Works Directive; mutual recognition with verification in Marrakesh; localization approaches in EU SatCab directive, EU Portability regulation, and DSM Directive. But attempt to enlarge localization to other aspects of © ran into hard wall in EU. Unification of exceptions is unlikely to do the trick.

Take some lessons from preexisting examples and models, recognizing that libraries and archives can’t move international change forward on their own.

Peter Yu (co-author Matthew Sag), The Globalization of Fair Use Standards for AI

Copyright Office is interested in whether there will be globalization on training and copyrightability. Different reasons for divergence: author’s rights/©; civil/common law; developed/developing. We look at training AI regimes. Group one: fair use. US, Israel (ministry of justice opinion embracing training); Liberia, Malaysia, South Korea, Sri Lanka, Taiwan. Canada is somewhat similar to US. Group II: TDM exceptions: Japan, UK, EU, Singapore, possibly China. Japan was earliest, in 2009. Not using for personal enjoyment = exception the broadest we’ve seen. Singapore is newest: computational data analysis (with limits to prevent excess dissemination). Interestingly, Singapore also has fair use.

Group III: Others, like China’s 3d amendment to Chinese copyright law. “Other circumstances as provided by laws and administrative regulations”—allowed limitations and exceptions. Seems to move towards an open-ended system. We’re waiting for implementing regulations.

UAE: efforts with Falcon AI. Law is very far behind; closest we’ve found is educational use.

No consensus. Emerging equilibrium. The underlying grundnorm: idea/expression distinction.

Countries do want to move forward; incentives to allow creation of systems in US/China/EU race. Gowers Review of IP—not adopted! But they believe that fair use would be better for UK to advance tech development. US dominance in AI investment and development.

US developments on deepfakes also matter—new ROP protections could change the environment. AI Act/not clear whether opt out or remuneration will proceed in EU. Also EU is trying to reach out with its regulation by being indifferent to jurisdiction in which AI was trained.

TDM/fair use is the way to go. There won’t be major international disputes on input end, but different modalities for implementation. Non IP issues will affect © development. Changes are likely to be incremental.

Hughes: you assume it’s fair use in the US. But we don’t know. Isn’t the real lesson that law doesn’t matter in the face of all the venture capital investment? We may never get a fair use determination b/c OpenAI and Google are doing deals with Reddit, and maybe the cases will settle.

A: we do think there’s enough case law for fair use, but what’s unclear is the interface b/t input and output.

AI and Authorship

Margaritha Windisch, Unveiling the Bond Between Artists and Their Work: A Vignette Study

Personality rights theory of ©: bond between artist and work/work is part of the personal sphere of the creator and part of their personal identity. Authors’ rights are designed to protect that intellectual and emotional bond. Strong emphasis on moral rights in continental Europe.

Fascinating empirical study among European visual artists randomly assigned to use traditional materials, digital drawing, or AI prompts. Their connection to their work felt significantly lower with AI prompts (somewhat lower with digital drawing but not nearly as much).

Time spent on creating matters to bond, but even controlling for that there is an effect. Whether their personality is reflected in the work/whether they had freedom of choice in decisionmaking—factors relevant to EU law; significant decrease for AI condition—whether they perceive the work as their own intellectual creation is much lower v. physical or digital creation. Differences b/t German-speaking and French-speaking participants as well!

There’s no threshold for the bond b/t artist and work; what’s the right policy to protect authors when there is a low bond? UK provides for © protection for computer generated work for 50 years, but no moral rights. The author is defined only as “the person by whom the arrangements necessary for the creation of the work are undertaken.”

[I’m really interested in Kant’s disrespect for visual art in initial arguments about authors’ rights, which is right there in the name (it’s not artists’ rights). Somehow people just skip past the medium in modern moral rights argument—but how and why?]

Jessica Silbey: We have bonds to lots of things in our lives that we didn’t create or don’t own.

Josh Sarnoff: rewarding for the bond seems to be rewarding the wrong thing—acquisitiveness or ownership desires.

A: no, the bond develops through the process of creation—what makes it difficult to produce is what develops the attachment to the work.

Q: consider storyboarding for a movie—playing with AI might be different.

A: agreed. Visual art has a tradition and might well be different for works that have had tech in the creation process from the beginning.