Friday, April 07, 2023

27th Annual BTLJ-BCLT Symposium: From the DMCA to the DSA—A Transatlantic Dialogue on Online Platform Liability and Copyright Law

Tutorial “The EU Digital Services Act – Overview and Central Features”

General DSA Architecture and Approach     

Martin Senftleben, University of Amsterdam

Formally, the safe harbor system is still in place for mere conduit, caching and hosting services for third-party information they transmit and store. For ©, voluntary own initiative investigations done in good faith shouldn’t lead to more liability. More intermediary services are covered under DSA. Points of contact for authorities/legal representatives are existing obligations but new are terms of service/content moderation policies and tools disclosure requirements, including algorithmic decision-making and human review, as well as rules of procedure of internal complaint handling. How much information must be disclosed? Must be balanced against trade secrets. Another new obligation: diligent, objective and proportionate application to safeguard fundamental rights. Obligation is on the service provider.

Transparency: reports on content moderation, including judicial orders, notices, number and type of content moderation, use of automated means and related training, and number of complaints.

Special rules for hosting providers, including online platforms; online platforms, including distance contracts, and very large online platforms and very large online search engines.

What about the copyright interface w/ the DSA? Art. 2(4): this is all w/o prejudice to rules laiid down by the © laws. So on top of all those distinctions, there are OCSSP obligations from Art. 17: online content sharing services. So where the © regime is silent, use the DSA. And there’s where the safe harbors have been gotten rid of: using the Art. 17 rules of required licensing—there’s direct liability if you’re unlicensed, but you can decrease the risks by using filtering to police the borders of your licensing deals.

OCSSPs are a controversial category in itself. YouTube is an OCSSP, but other than that much is unclear—are all different types of social media OCSSPs? We don’t have definite caselaw.

DSA requires VLOPs and VLOSES to have reasonable, proportionate and effective mitigation measures tailored to specific systematic risks w/particular consideration to the impacts on fundamental rights. Art. 17 is more specific and thus a further addition to DSA requirements.

“Sufficiently substantiated notice from the rightsholders” and rightsholders should “duly justify the reasons” for their requests—opening for more specific requirements like Art. 16 of DSA which say that notices have to be sufficiently precise/adequately substantiated; so to statements of reasons must be “clear and specific.”

DSA Art. 23: suspend, for reasonable period of time, and after warning, users who have violated rules; so too processing of notices and complaints by complainants that frequently submit notices or complaints that are manifestly unfounded—not present in © framework so additive to it. Also Art. 22 requirement to ensure that notices submitted by trusted flaggers w/in their designated area of expertise, should be processed w/o undue delay. Must comply w/quality standards—status awarded by Digital Services Coordinator of Member State if flagger has particular expertise of detecting/notifying illegal content; is independent of any provider of online platforms (but does not need to be independent of © rightsholders); and carries out its activities diligently, accurately and objectively.

DSA also requires VLOPs and VLOSEs to get annual independent audit opinions, reporting to Commission, potential for fines, so they can’t just wait for court cases to identify whether they’re doing it right.

Fred von Lohmann: For enforcement: other than the intervention of the Commission through the audit process—how will enforcement be done? Private rights of action?

A: DSA enforcement is very much by the Commission itself in his understanding. Drafter of DSA, Irene Roche-Laguna: it depends on the kind of platform—hosting service, platform, or VLOP. EC has exclusive powers for enforcement of obligations that apply only to VLOPs/VLOSEs—auditing/risk assessment. But content moderation requirements for platforms in general are for country of establishment (if in Europe) and where the legal representative is (if not). If a platform neglects notices it becomes a systemic issue; the Commission can jump in. [I wasn’t sure whether she meant that was true even for non-VLOPs.]

Q: can a recipient of the service receive money damages for violation of these obligations? Is that governed by Member State law?

A: In DSA 54, there is an avenue for private actors to bring damage claims to complement audit-based or other executive activities.

Pam Samuelson: will audits be publicly available?

Roche-Laguna: Audit reports will be published w/transparency reports every 6 months, and audit implementation reports have to be published. They will have a bit of time to respond.

A: given the transparency requirements, the audit report should be able to provide more insights, but we don’t know for sure yet.

Q: what makes an online platform?

A: central aspect is dissemination of information. [Husovec’s piece says Blogger and Wordpress aren’t platforms, which seems bizarre to me.] What is a public? It may not have to be a large public. In © we accept that dissemination to a circle like 12 or 50 people can already be a public, so would that also apply under the DSA? The Court of Justice could set a relatively low threshold.

Roche-Laguna: there is a gray area. A hosting service that disseminates to the public as ancillary feature, like a comments section on a website, that wouldn’t be a platform. But we have questions about messenger services with open chats/groups—100,000 people in chat. Or music service that has licensed content and user-uploaded content: what is the platform part? [That doesn’t seem to answer questions like, is email dissemination to the public?]

Q: do platform rules have to be observed for the ancillary service? It’s a lawyer’s paradise. [Roche-Laguna is laughing but I’m not finding it funny that they actually can’t seem to answer those questions.]

Rules for Hosting Providers, Online Platforms and Very Large Online Platforms  

Martin Husovec, London School of Economics

DSA is about content moderation, construed very broadly. Not just removal of content as w/© infringement but decisions about violations that aren’t based on legality but are purely contractual—FB and breastfeeding images; suspending individual users. Also about institutions lower in the stack like app stores. Hiding/demonetizing content is also covered. DSA covers almost everyone, though advertising providers and payment providers are touched only tangentially. Infrastructure providers like access providers, transit services, DNS/VPN services, domain registrars and registers, generally have very few obligations other than transparency. But distribution and content layer have more obligations.

Part of the second generation that replaced the first generation like section 230 designed to create breathing space for speech and industries; focus on is regulation of risks posed by services.

For content moderation: fairness in design ex ante, due process in implementation, transparency, and risk management. Due process constrains the moderation of both illegal content and contractual breaches. Risk management for very large players: required to think about product design and operations; w/some risk management obligations for smaller platforms.

Online platforms are covered if 50+ employees/EU turnover of 10 million euros. VLOPs/VLOSEs: Alphabet, Microsoft, Meta, Bytedance, Snap, Pinterest, Twitter, Amazon, Wikipedia as the only nonprofit.

Decide what rules are; open to notification from third parties about content that might be illegal or violate terms & conditions; make decisions (hosting services and any-size companies); allow appeals/internal contestation (midsize and above); allow external contestation by users (VLOP/VLOSE); transparency. Any content restrictions has to be codified in terms and conditions, and decisions must be made on the basis of the rules. But the rules must be diligent, objective and proportionate, including in their design according to the recitals. What does it mean for design of rules? His take: only extreme outlier policies would have an issue.

Core of DSA: tries to discipline providers in how they make decisions. They must issue a statement of reasons, including for visibility, monetization, etc. Specific explanation of reasons required, useful enough to allow user to argue against them. They can be automated, but DSA says you need to provide a free opportunity to appeal where a human has to be present in some step of the process. Internal appeal must be easy to access and user-friendly; not solely automated; must occur in timely, diligent, and objective manner.

Who can complain or appeal? External players: trusted flagger, regular notifier, content creator, NGOs.

Possibility of external dispute resolution/ADR: Regulators certify entities independent of platforms and users; FB’s Oversight Board is not independent. Content creators and notifiers and reps can use the option. ADR provider is complainants’ choice; no need to exhaust appeals. ADR decisions are nonbinding; platforms must engage in good faith. Plaintiff compensates complainants who win (fees and possibly costs). Complainants that lose pay their own fees and costs. Trying to improve quality of decisionmaking within company.

The proceduralist approach constrains implementation more than rule formulation. Consider: for 5 EUR a month, you can say whatever you want on the service as long as it’s legal in your country—disinformation, nudity, etc. all ok. Everyone else is moderated on ToS violations and illegality. Is this a violation of art. 14(1)? No if disclosed. Art. 14(4): probably no—paying 5 EUR to be unmoderated doesn’t seem disproportionate. What if you have a list of VIPs whose content is not moderated at all b/c they are leaders of countries? Again, 14(1) is ok if properly described. But 14(4) would be a problem in his view due to the impact of illegal content.

Open issues: big guys can easily automate statement of reasons; what about small providers? Licensed content moderation solutions from third parties? What about users of services, like newspaper’s FB page—are they considered to be a separate entity for purposes of people whose content they moderate? How will there be standard transparency database if different people are sending their reports in nonstandard format and with anonymization?

Risk management for VLOPs/VLOSEs will likely not be subject to private enforcement—mostly extended or intensified reporting obligations; researcher data access; unique obligations for profiling-free choice on recommender systems (whether organic or advertising), or advertising archives. Subject to regulatory dialogue w/whole community, including national regulator, NGOs, researchers, where main thing is to figure out what companies are doing w/r/t certain types of risks on their platforms. Dialogue b/c of opacity of system and info asymmetry: regulators can’t instruct on what to do w/out knowing what’s going on. Must assess systemic risks stemming from design or functioning of services including algorithms and use made of their services, as well as back end governance, taking into account the risks’ severity and probability. Risks include: illegal content; fundamental rights; public security and elections; health and well-being (including gender-based violence, public health, minors, physical well-being, and mental well-being). Nothing of concern to civil society is left out. If a social network adopts ChatGPT into design, that becomes regulated along with anything else the platform uses.

Metaphor: authorities can partly restrict how and when protest activities take place (streets, hours, use of amplification tools) and take measures to prevent harm to protestors or others (e.g. by boosting police presence) but can’t select speakers or dictate content.

Q: why is business to business included here? Why can’t Amazon say “one strike and you’re out” to a business?

A: b/c the design was about risks of hosting services, not just harms to individuals or individuals as consumers. Advertising marketplaces are available to the public.

Q: what about risk mitigation for humans conducting content moderation/labor harms?

A: sure.          

Interplay with OCSSP Rules in the Directive on Copyright in the Digital Single Market

João Quintais, University of Amsterdam

Background of lots of attempts to interpret previous rules under InfoSoc Directive. Primary liability is harmonized. But secondary liability was mostly unharmonized; ended up with mostly a notice and takedown regime. Court of Justice expanded right of communication to the public to the point where there was a Q of whether YT’s own services triggered direct liability. Poster child was YouTube/claims of “value gap” (sigh).

Six states haven’t met the implementation deadline; still waiting to figure out what the rules actually should be. Poland decision: CJEU said Art. 17 was ok but had to give due regard to uploaders’ interests. OCSSPs cover UGC platforms with large amounts of works that organize and promote them and have a commercial/competitive effect. Exclusions: encyclopedias, ecommerce, B2B/cloud hosting: Wikipedia and GitHub/ArXiv.org, Skype, Dropbox, and eBay are excluded—partly down to who had good lobbyists. But you are still covered by InfoSoc directive plus DSA. Startup provisions: under 3 years and 10 million Euros, only notice & takedown, but if above 5 million visitors also notice and staydown. Not much there. Could be excluded from DSA but covered by Art. 17.

A bipolar copyright system/an employment program for EU lawyers. You might have to look service by service to figure out whether you are an OCSSP. Etsy? Wordpress?

Non-OCSSP: default no direct liability/hosting safe harbor and moderated duties of care, based on YT case and national law. OCSSP: default is direct liability w/exemptions tied to best efforts licensing and filtering.

Most online platforms for DSA will be OCSSPs so you have to figure out what applies—DSA may cover things only partially, and it’s not very clear/depends on your normative preferences. Bonanza for lawyers! If you’re a non-OCSSP, DSA will probably apply to almost everything.

Von Lohmann: Secondary liability seems to have disappeared. It’s not just OCSSP/non-OCSSP—there’s no harmonization for secondary liability, and safe harbors are interpreted as not precluding injunctive relief in most jurisdictions—so don’t you also have an entire category of secondary liability injunctions that are outside of both of these regimes or they displaced by DSA/Art. 17?

A: you are right, there might be that category. YouTube case is designed for YouTube, but is now covered by Art. 17: need to apply that to other platforms, and how to do so is unclear. German courts are deciding how to modulate duties of care and will continue to do so unless ECJ tells them to stop. In all cases, you assess liability also on the basis of compliance w/duties of care, but DSA duties are different: obligations regarding user-uploaded content are about your role as platform.

Comment by Senftleben: may not be as bad as that—the CJEU has extended primary liability so far that there’s not much room left for platform secondary liability. The crucial question, and an open one, is what can be expected from a reasonably diligent operator in this situation—will this be influenced by the new DSA duties?

A: court will likely look at whether this is a good faith player—will reason backwards to find no liability if so. The “good guy” theory of EU copyright law: court finds a way around liability for a good faith player, and a way for direct liability if there’s not. You can’t really get from the text to the outcomes; the court just makes up a bunch of conditions.

Comment: primary and secondary liability are no longer distinct categories. Now all under the umbrella of primary, but if you can’t get authorization you have to filter, so the duty of care is now embedded in primary liability.

Beyond Copyright Infringement: DSA Review, Moderation and Liability Rules Compared to Previous National Review and Takedown Approaches for Illegal Content

Matthias Leistner, LMU Munich Faculty of Law

Need to know whether national rules are preempted, and they also provide evidence of experience that might be useful for implementing DSA. There were differences in applying and enforcing rules; national authorities might have leeway in applying DSA rules, which might lead to forum shopping about where to have a company seat.

German NetzDG regulates allegedly criminal content on social media (obvious v. non-obvious are treated differently). Austria has a similar law; France’s law was partly invalidated b/c of extremely short blocking periods for certain content, but it also has a law against fake news in 3 months before a national election; Italy has a law against “cybermobbing” and dealing with parents/minors; Baltic states have their own.

NetzDG has limited scope: social networks, video sharing platforms v. DSA’s comprehensive/four-tiered approach.

NetzDG: catalogue of substantial/hard-core criminal offenses, continuously extended v. DSA’s all illegal content including minor violations (including consumer protection law), plus special notification duties for criminal acts etc.

NetzDG: persons affected in their rights can notify; DSA: everyone can notify.

NetzDG: expeditious blocking: 7 days or 24 hours in obvious cases; DSA: expeditious/without undue delay.

NetzDG: redress for both sides; another human in the loop; DSA: role of complainant surprisingly unclear, no standing for the other party in the complaint.

NetzDG: No further prioritization b/c of limited scope; DSA: system of trusted flaggers given priority; dynamic adaptation through provisions on misuse and suspension.

NetzDG: limited and crisp; DSA: wide and wobbly.

Provisions on transparency of user contracts are similar to German case law; reasonable and cost-efficient b/c contracts exist anyway. But may not be easy to implement b/c some transparency obligations relate to what the algorithms actually do. [I think this may undersell the difficulty of figuring out what is, for example, abusive or hateful speech, especially for non-dominant communities.] Protection of trade secrets is also a huge issue.

Most probably, NetzDG will be repealed in 2023, but does the DSA entirely preempt Member State laws? Depends on the harmonized subject matter—access for research purposes? Research access for other public policy purposes?

Was there overblocking b/c of NetzDG? No empirical evidence under the procedural mechanism, but a backlash to unjustified blocking under different user policies. Was preventive blocking under user policies/community standards perhaps indirectly due to NetzDG? Avoiding cumbersome statutory mechanism by overcompliance? Hard to assess w/o access for research.

Indirect regulatory effects have probably worked quite well w/o overblocking; proposed self-regulation boards for notice and takedown don’t seem to have worked. Platforms didn’t want to cooperate and platforms have different policies w/regard to user communities.

CJEU has responsibility for hard questions, along with national coordination; resources and coordination are issues and GDPR experience is disheartening though antitrust experience is better.

Q: regulations require clear and understandable rules in terms that are neither clear nor understandable: how are platforms going to establish such rules for billions of pieces of content?

A: it’s not new that the legislator can do things that a private entity can’t. German FB case established certain basic standards for typical user policies that would be easy to comply with. There must be certain reasonable rules on what is allowed and what isn’t. DSA requires examples for rules of what is prohibited and what is not. Pretty much about procedures. [But people will always be able to argue that their slur is closer to the not-prohibited side than the prohibited side, if their slur isn’t in the list of examples. This reassurance might work for legal/illegal but it is not helpful at all for lawful but awful.] It’s about the bad guys with self-contradictory community policies that block without reasonability.

Midsize platforms can game the system by choosing a friendly jurisdiction like Ireland. Carveouts don’t protect enough small providers and will be problems for Baltic startups.

Q: it’s nice to say that Germany doesn’t require too much detail but this is a new regime, and France is going to be different.

A: agree to a certain extent; also important that some standards are new—certainty will take 10 years and that’s not great. But the transparency obligation doesn’t come from a vacuum—consumer protection law/transposing concepts of standard terms to the platforms [where the subject matter of the transaction, the user’s speech, is pretty different!]. While we were discussing DSA, it made sense to object and discuss our concerns, but now we have it, so it’s time to think about ways to make it workable to the extent possible. The only way forward is to implement it.

Daphne Keller: share concerns about private litigation in different member states, lack of harmonization. How many states might be interested in allowing this?

A: We had this problem before, w/27 different standards—how many of them really matter? Here we have to distinguish b/t public and private enforcement. Public enforcement, the Commission will matter/establish standards that might trickle down from VLOPs to smaller platforms. Germany, France, Netherlands, Scandinavian countries will matter.

Keller: DSA is silent on question of what injunctions can say.

A: private enforcement: new provision on damages and injunctions might be possible; would have to look at different states’ laws. Largest dangers are Germany and France national unfair competition laws, where competitors might be able to sue—real problem w/GDPR that we are still grappling with. Practically limited to Germany, France, maybe Austria and some others. So far, Austria legislated specifically that compliance obligations don’t qualify for injunctions, period; caused quite a number of problems in the market.

Free Speech Challenges and Potential Risk Reduction in the DSA

Eleonora Rosati, Stockholm University

Balancing freedom to conduct a business and the need to protect recipients of services through transparency. Enhanced obligations for VLOPs and special provisions for micro/small enterprises. Art. 17 tries to protect users with certain exceptions.

Copyright and free speech: EU Charter of Fundamental Rights is a primary source of EU law, and recognizes IP within right to property; also recognizes free speech. Also recognizes freedom of artistic expression, freedom to conduct a business, respect for private and family life, and protection of personal data—not just integration of markets but integration of morals. Balancing framework, though how the balancing is to be done is the key. Level of protection for © needs to be high—what does that mean? High level of protection for whom and of what? InfoSoc Directive refers to authors, performers, producers, and culture industry alike, as well as consumers and public at large. There are other kinds of rightsholders. ECJ has been clear that high level of protection does not mean highest level of protection; the goal is a fair balance of different interests. Increasingly characterized by a fundamental rights discourse.

Many concerns about free speech from Art. 17, resulting in action by Poland:. Resulting judgment: ECJ didn’t say Poland was wrong that it restricted free speech—free speech is not absolute and can be restricted under certain conditions. ECJ said there were enough safeguards: filtering tools must be capable of adequately distinguishing b/t unlawful and lawful content; users must have rights; rightsholders must provide relevant and necessary information; no general monitoring obligation; procedural safeguards must exist; and there should be stakeholder dialogue and fair balance.

Inconsistent approaches at the national transposition level, so Art. 17 won’t be the stopping point. Also, lex specialis/lex generalis relationship is likely not to be frictionless. CJEU will have to smooth things out.

Comment: if you overblock, all you face is user complaints. If you underblock, you get sued and it’s costly. That’s where the balance is.

A: mostly the remedy for overblocking is to let the thing up; though national law might have some other remedies.

Comment: German law has a sort of must-carry obligation if a user asserts the applicability of an exception. But it’s the minority approach.

A: Italy does the opposite—if there’s a complaint, content must be disabled during the dispute. From platform point of view, you may have to divide treatment of Germany and Italy/geoblock.

Q: is Italy compliant with the Poland CJEU decision?

A: neither Italy nor Germany is—you can’t have a blanket approach. [I don’t understand what that means—I would appreciate hearing what a provider is supposed to do.]

Exceptions and limitations might mean something different in terms of whether they establish user rights—or maybe not! The new data mining provisions are characterized differently from each other, but not clear whether that makes a difference to whether, for example, one is entitled to remuneration and one isn’t.

Comment: if you take auditing seriously, maybe there can be changes/real examination of the practices.

Von Lohmann: but who selects and pays the auditors?

Comment: they’re supposed to be independent!

VL: that’s not the question.

Q: will this get rid of small platforms b/c all rights have to be equally honored?

A: it depends. ECJ is adamant that fair balance depends on circumstances of the case. Some rights might take precedence on a particular platform; not all content is created alike. There might be situations in which certain concerns override other types of rights. France would maybe answer your question in the affirmative, but not all would.

Comment: situating different categorization options under DSA, might see cases that the more you diversify the types of content you host, the more obligations you have.

Q: for US lawyers, we look at the Texas and Florida laws that are clearly politically motivated attempts to suppress content moderation in favor of a political agenda. Hard time understanding how that’s going to be handled in Europe, but the language of nondiscrimination suggests that it can and will happen in Europe too, which also has culture wars [and backsliding democracies].

A: indeed, the concerns around © are also in line with those concerns. The discourse is broader than ©--Afghanistan papers case, where German gov’t attempted to repress publication of military reports about operations of German army in Afghanistan. Used © as a tool. Advocate General and ECJ emphasized that © can’t be used to suppress free speech and one should be careful to limit © to its purposes.

No comments: