URL originale : https://books.openedition.org/pressesmines/3701
Managing internet risks
Texte intégral
Designing an impact assessment for internet regulation
1This book addresses the question of how regulatory impact assessments should be conducted in the context of internet regulation, focusing in particular on cost-benefit analyses. What would a cost-benefit analysis look like for measures designed to protect citizens against harmful internet content? How can protection of fundamental rights and the internet ecosystem be integrated into a cost-benefit analysis?
2Considerable science has been devoted to regulatory cost-benefit analyses, particularly in the context of environmental, health and safety regulations in the United States. The OECD and the European Commission have likewise endorsed thoughtful cost-benefit analyses as a precondition for any new regulatory measures. And yet cost-benefit analyses are rarely if ever done for laws regulating internet risks. Why focus on internet regulation? Like environmental risks, digital risks can trigger overreaction by policymakers eager to show that they are doing something about a problem created or amplified by new technology. The problems are generally real, but policymakers almost never analyze the problems and the scope of possible solutions with scientific rigor. Why should they? Politicians and regulators are generally rewarded for creating new laws and regulations, not for doing nothing. Yet a rigorous cost-benefit analysis may show that doing nothing is better than regulating. Regulatory over-reaction occurred in the United States in connection with certain health, safety and environmental risks in the 1970s and 1980s. A number of scholars highlighted the huge discrepancies in costs of many of these well-meaning regulatory measures. Some measures saved many lives at little cost; others saved few lives at astronomical cost. The purpose of the cost-benefit analysis is to highlight these differences, so that regulators can make rational trade-offs.
3This book examines the most common form of internet risk – the risk of harmful content – and the most common form of internet regulation – regulations targeting internet intermediaries – and proposes a cost-benefit analysis for considering these risks and regulatory responses. The cost-benefit analysis is part of a broader impact assessment, the purpose of which is to increase the quality of regulatory measures. The quality increase would occur at several levels. First, the impact assessment will decrease the likelihood of parliaments and regulators rushing into regulatory solutions that are ill-adapted to the fast-moving internet ecosystem. The impact assessment would reduce the frequency of these errors by imposing more rigorous fact-finding, a more thoughtful definition of the outcome that the measure hopes to achieve, and ways to measure success. The assessment would require a systematic consideration of costs, including indirect costs that are largely ignored in impact assessments today.
4Second, the impact assessment will enhance regulatory quality by fostering competition and criticism between regulators and institutions. Systematic comparison and criticism based on a standard methodology would improve quality by permitting best practices to emerge. Today, measures are adopted in different “silos”. The silos are based on country (e.g. Italian measures versus French measures) or based on content policy (e.g. copyright infringement policy versus right to be forgotten policy). One measure may be adopted to limit access to child pornography, another adopted for illegal gambling, and another yet for online copyright infringement with little or no coordination between the three approaches, and little debate on which approaches are most effective. The methodology would permit comparison and learning across silos.
5Third, impact assessments would increase quality by enhancing a measure’s international legitimacy. If the methodology proposed in this book is recognized by international institutions such as the OECD and the European Commission, regulatory measures that emerge after a rigorous impact assessment would benefit from a favorable image internationally – a form of international quality label. The international quality label would signify that the country adopting the measure did its best to apply good regulation principles such as proportionality and respect for the internet ecosystem when adopting the measure. The impact assessment would presumably be available for public scrutiny, permitting international observers to verify that the methodology was applied correctly.
6The proposed methodology will not yield a single good answer for any policy problem, but will permit regulators to rank alternatives based on their relative impacts on factors that are important in the context of internet policymaking, including fundamental rights. The final decision on which regulatory option is best adapted to a given situation will in most cases remain political. The political decision may result in a choice that is not necessarily the optimal choice under the methodology. In that sense, the methodology is an input to the decision process rather than a decision rule of its own (Posner, 2000).
7Nevertheless, my hope is that a standard methodological benchmark will lead to measures that are more consistent, more proportionate, and easier to understand for stakeholders, and that the methodology will avoid policymakers exaggerating internet-related risks and reinventing the wheel each time a new internet regulation is put forward. The methodology would also permit Europe to propose a standard that would justify regulatory intervention in certain cases, while permitting Europe to distinguish its approach from that of other less democratic countries that use internet intermediaries as tools for censorship. Interfering with internet content is a form of non-neutrality. Once regulators start down the road of internet non-neutrality, it can be hard to stop the movement. A uniform methodology would help define limits, and distinguish European measures from measures implemented in other less democratic countries. A methodology would contribute to defining what is meant by the “open internet.”
8The main difficulty we will encounter is attempting to integrate fundamental rights into the cost-benefit analysis. Many of the benefits flowing from a content policy (promotion of culture, protection of privacy or human dignity) are hard to quantify, let alone convert into monetary units. Many of the indirect costs of using internet intermediaries as proxies to enforce content policies will also be hard to quantify: limitations on freedom to access information, threats to privacy, harm to the internet ecosystem. But as we will see, there are ways of approaching the problem of quantifying these hard-to-quantify values.
Bringing smart telecommunications regulation to the internet
9Part of my career has been devoted to assisting telecom regulators in Europe justify the imposition of new regulatory measures on telecommunications operators. The task is difficult because the European directives on electronic communications impose a strict methodology that regulators must follow before they can propose new measures. Regulators must conduct a market analysis, identify one or more actors holding the dominant position, demonstrate that a market failure creates durable barriers to entry for competitors, that technological and market evolutions are not likely to cure the problem without regulatory intervention, and that competition law is not sufficient. Regulators must also show that the measure they propose is proportionate, causing the least intrusion possible while achieving the desired outcome. In some countries, regulators must present several alternatives and choose the one that is the least burdensome while still attaining the desired objective. The regulatory proposal must take into account the principles set forth in the Framework Directive1 on electronic communications: encourage competition, investment, respect for technology neutrality. Regulators must submit the proposed solution to public consultation and then to a special task force at the European Commission, which has the power to request changes and in some cases veto the measure. Finally, the measure must be reevaluated regularly to make sure it is still fit for purpose. Regulations that are no longer needed must be withdrawn.
10This methodology contrasts with the total lack of methodology for regulatory measures designed to limit access to content on the internet. Measures adopted to fight online copyright infringement, hate speech, child pornography, and privacy violations often involve internet intermediaries, including telecom operators. Yet those measures are adopted without the analytical rigor that applies to European telecommunications regulation or to health, safety and environmental regulation in the United States. The potential adverse effects of the proposed measures are not studied in detail. The impact assessments omit major considerations relating to effects on fundamental rights and adverse effects on the internet ecosystem. There is no peer review system, and no system to remove regulations that are no longer fit for purpose. If these measures were presented in a context of telecom regulation, many of them would fail the strict scrutiny imposed by the European directives.
11The study of telecommunications regulation and the literature surrounding cost-benefit analysis in the United States led me to the idea of this book. Would it be possible to take the analytical tools applicable to European telecommunication regulation and United States cost-benefit analyses and transpose those tools to the field of regulating access to internet content?
12The question of limiting access to content is more complex than telecommunication regulation because the objectives pursued by policymakers include protection of fundamental rights, protection of children, and national security. The objectives of telecommunication regulation consist principally of achieving effective competition. The question of access to illegal content on the internet is also more complex because of the number of different internet intermediaries involved. In addition to telecom operators, search engines, app stores, social media, advertising networks, and payment providers may be called on to assist in enforcing a given content policy. However, the complexity of the problem is all the more reason to use analytical tools. The European methodology for telecommunication regulation is not perfect, but it requires regulators to ask key questions before they act: does this market failure really require a regulatory response, or is the market likely to deal with the problem on its own? Is the regulatory measure we propose the least intrusive among the available alternatives? What are the potential side effects of our proposed measure on competition, innovation and investment? These questions must be seriously analyzed in any proposed regulation of telecommunications in Europe. Cost-benefit analysis in United States regulatory policy requires that policymakers precisely define the desired outcome and develop tools to measure the outcome. In this book I will show that the same questions (and others) should be asked and analyzed in the context of regulations designed to limit access to harmful content on the internet.
The challenges of regulating access to content on the internet
13Publishers of content and services on the internet are often located beyond the reach of national courts and police. Because the publishers are beyond reach, lawmakers and courts tend to look to internet intermediaries located within national borders as proxies to help apply content rules (Noam, 2006; Lichtman & Posner, 2004; Lescure, 2013). The internet intermediaries may be ISPs, payment providers, search engines, social media, app stores, domain name registries, browser publishers, or advertising networks (OECD, 2011).
14To illustrate the difficulty facing regulators, let us use the example of hate speech: French law prohibits content that promotes racial hatred or anti-Semitism. In the United States, many forms of hate speech are protected by the First Amendment of the United States Constitution. Hate content that is published on a website based in the United States is instantly accessible by citizens of France. French victims of hate speech can bring an action before French courts and then try to obtain enforcement of the action in the United States. However, enforcement of a French judgment in the United States will be long and costly, and a United States court will not necessarily enforce a French decision that potentially interferes with United States constitutional principles. Even if a United States court were to grant enforcement, the publisher of the content could easily move to another country and start its website again.
15As this example shows, going after the source of the offending content is difficult and in some cases impossible. That leaves the option of seeking enforcement against technical intermediaries located in the country where the harm is caused. Take our example of the hate speech sites based in the United States. A number of options are available in order to make the hate speech sites less available to French citizens. Internet access providers in France could block access through various kinds of filtering. A search engine could make the site less visible on search results for French users. If the site collects payments from French users, payment providers could be called on to block payments. Browser software could be configured to make access to the site difficult. Advertising networks could be called on to block advertising. If the site uses a.fr domain name, the domain name could be seized in France.
16All of these techniques can potentially be used to limit French citizens’ access to an offshore hate speech site. However, all of these measures have potentially grave side effects. Technical filtering can block the targeted hate speech but can easily block other perfectly legal content. Filtering can create significant costs for internet intermediaries, can interfere with principles of net neutrality and threaten privacy. Moreover, many measures may prove ineffective, and/or encourage users to use encryption and dark networks to avoid detection, which creates other problems for law enforcement authorities.
17National regulators may even be tempted to make their blocking measure apply to the entire world. In Europe, individuals have the right to request search engines to remove certain harmful search results, even though the content revealed by the search is not illegal. Some European privacy regulators ask search engines to apply the blocking measure to search results worldwide, arguing that the victim is entitled to effective protection even if the search is conducted outside Europe. If applied to search results in the United States, the blocking measure would block content protected by the First Amendment of the United States Constitution. The measure would also set a precedent for other governments who may want to silence dissent by asking a major search engine to apply global censorship. Like a potent medicine, measures taken by internet intermediaries, whether on their own or under government constraint, can have dangerous side effects.
18The principles of the open internet and net neutrality prohibit interference with the free flow of content, applications and services on the internet. The open internet creates numerous economic and social benefits (OECD, 2016). Actions by internet intermediaries, whether government-imposed or voluntary, will necessarily affect the open internet. Interferences with the open internet are tolerated if they are part of “reasonable network management,” that is, if the measure is intended to address a legitimate objective, and the means used to attain the objective are proportionate (Sieradzki & Maxwell, 2008). Net neutrality so far applies only to ISPs, but its principles can in theory be extended to any kind of internet intermediary (ARCEP, 2012).
19Net neutrality has at least three objectives: to prevent anti-competitive conduct by last-mile ISPs, to protect freedom of expression and to protect the borderless and end-to-end character of the internet (Curien & Maxwell, 2011). Measures that erect gateways designed to protect national content rules on the internet constitute a serious threat to the open, global character of the internet. Advocates of net neutrality also fear that if democratic countries begin to impose non-neutrality to achieve content objectives, other less democratic countries will follow suit, with measures to enforce political censorship or religious doctrine (OECD, 2011).
Current regulatory approaches are uncoordinated, with no guiding methodology
20Today many systems exist to enforce content rules on the internet, but they are developed on an ad hoc basis to deal with specific problems (Mann & Belzley, 2005).
21Courts, regulators and lawmakers adopt different approaches for online copyright infringement, illegal gambling, hate speech, child pornography, right to be forgotten, and terrorism cases. There lacks a reference methodology against which to measure these initiatives.
22To illustrate the point, below are examples of French measures adopted to address different content issues. These examples show the diversity of approaches used even within a single country.
Right to be forgotten
23The right to be forgotten permits an individual to ask a search engine to remove certain search results that appear when someone conducts a search using the individual’s name. The underlying content is not illegal. If it were, the individual could ask for removal of the content at its source, using legal claims such as defamation or invasion of privacy. In the case of the right to be forgotten, the original content (e.g. a newspaper article) is legitimate, protected by law, and remains available on the internet. The individual simply wants the content to be less easy to find in search results because the content is old and harms the individual’s current life.
24The right to be forgotten flows from a court decision interpreting the broad provisions of the European Data Protection Directive 95/46/EC, and in particular the provisions that guaranty each individual a right to object to processing of his or her personal data.
25For the so-called “right to be forgotten,” the French data protection authority, the CNIL, has assumed the role of dispute resolution body in situations where claimants are not satisfied with the solution proposed by a search engine. In its dispute resolution capacity, the CNIL relies on the European Court of Justice’s decision in Google France vs. AEPD2 (the “Costeja” decision). The CNIL applies the principles of the Costeja decision to France’s Data Protection Act3, and then issues an individual decision ordering the search engine to remove a certain search result from searches made using the individual’s name. As indicated in the Costeja decision, the claimant’s right to be forgotten request must be balanced against the public’s right to have free access to information. If the relevant information is irrelevant, outdated and harmful to the individual, the CNIL would grant the request unless the public has a legitimate interest in having access to the information. This would be the case if the claimant were a public figure, for example. If the CNIL is satisfied that the balancing test comes out in favor of the claimant, the CNIL will order the search engine to remove the search results whenever an internet user conducts a search using the relevant individual’s name. The CNIL’s position is that the search engine should eliminate the search results from all searches worldwide, regardless of the country from which the search was initiated. The CNIL’s decision therefore would have extraterritorial effect, limiting the information that would be seen by an internet user in the United States, for example. In rendering its orders, the CNIL currently does not take into account in its balancing test the fact that the search engine’s action would likely impede access to content protected by the First Amendment of the United States Constitution. Similarly, the CNIL does not take into consideration in its balancing the precedential effect that a global order against the search engine could have for other countries who may also wish to apply their own content policies worldwide.
26The current “right to be forgotten” only applies to search engines. Other internet intermediaries are not affected. The publisher of the original content, and the hosting provider, are under no obligation to remove the relevant content because the content itself is not illegal. The right to be forgotten is therefore unique in that the objective sought is not to remove or block access to the original content, but instead to make the original content more difficult to find using certain search terms and a certain kind of internet intermediary.
Online copyright infringement
27France was the first country in the world to adopt a regulatory framework for fighting online copyright infringement using the so-called “graduated response” approach. Under the HADOPI4 graduated response regime, right holder organizations collect IP addresses of suspected infringers using peer-to-peer networks. The evidence is then transmitted to the HADOPI regulatory authority, who then asks the internet access providers to communicate the names of the subscribers corresponding to the IP addresses. According to HADOPI’s activity report for 2013 – 2014, 12,265,004 identification requests were sent in total to the internet access providers. Once the HADOPI receives the names of the subscribers, HADOPI can take three steps. First, the HADOPI sends an initial e-mail to the relevant subscribers informing them of their duty to ensure that their internet access is not used for infringing purposes, and reminding the subscriber of the existence of legal online offers. According to its activity report for 2013 – 2014, the HADOPI has sent out 3,249,481 first warnings. Second, repeat infringers then receive a registered letter from the HADOPI stating that the subscriber has been identified again as the source of infringing content, and that if the conduct does not cease the HADOPI may transmit the file to the public prosecutor for sanctions, which may include suspension of internet access. According to the figures published by the HADOPI in 2014, 333,723 registered letters of this type have been sent. For subscribers that continue to show evidence of infringing activity, the HADOPI then selects, in the third step, the files to be reviewed and may ask the relevant subscriber to participate in a hearing. The HADOPI can send the files to the public prosecutor if infringement continues.
28In addition to relying on the HADOPI graduated response system, victims of copyright infringement have successfully obtained French court orders to block access to streaming sites. Finally, the French Minister of Culture has nudged the principal French internet advertising players to agree to refuse to purchase advertising space from internet sites that manifestly promote illegal copyright infringement.5 The list of the sites affected by this measure will be put together by an industry coordination committee based in part on a list provided by the French police authorities. The code of conduct does not provide for any sanction against advertisers that violate the code. The French Minister of Culture also hopes that a similar code will be signed shortly by banks and payment service providers. The code would prohibit payment service providers from knowingly providing service to sites that promote copyright infringement.
Illegal online gambling
29France allows online gambling, but only with gambling service providers that have obtained a license. The licensing conditions are intended to protect individuals against the harms associated with addictive gambling, as well as to protect society against the development of organized crime around online gambling activities. The French regulations on online gambling are administered by a specialized regulatory authority called the ARJEL.6 French law gives the ARJEL authority to take measures to limit access by French users to unlicensed gambling sites. The ARJEL has authority to draw up lists of unlicensed gambling sites to which access should be blocked. The ARJEL then submits the list to a court in an ex parte proceeding. The court then issues an order requiring that French ISPs block access to the relevant sites by inserting an erroneous IP address for the site in the access provider’s local DNS server. The decree relating to ARJEL’s blocking authority specifically provides for DNS blocking.7
Child pornography and terrorist propaganda
30For internet content involving either child pornography or incitement to commit terrorist acts, the French police authorities are able to send blocking requests directly to ISPs without first obtaining a judge’s approval.8 The police must first attempt to obtain removal of content at its source through a request to the publisher and hosting provider, but if unsuccessful after 24 hours, the police may direct their request to local ISPs. The decree relating to blocking of child pornography or terrorist sites does not specifically refer to DNS blocking.9 The decree says that ISPs should block access to addresses “by any appropriate means”, and redirect visitors toward a website of the French police. The law refers here to blocking “addresses”, not to blocking “sites” or “content”, which suggests that ISPs would use DNS blocking rather than other more intrusive forms of blocking such as deep packet inspection. The French government must reimburse ISPs for the cost associated with the blocking measures.
31To compensate for the fact that no judge is involved in blocking decisions, the law provides that a person named by the French data protection authority will receive copies of blocking requests and can issue recommendations to the police authorities, or ask a court to intervene. The law relating to child pornography and terrorist site blocking also authorizes the police to address removal requests to search engines and directories.
32In addition to these regulatory measures, many internet intermediaries apply self-regulatory measures to facilitate the reporting and blocking of child porn sites. This is done through an international reporting network called “INHOPE” (www.inhope.org).
Hate speech
33French law prohibits content that incites racial hatred or anti-Semitism, as well as content that incites discrimination or hatred based on sex, sexual orientation or handicap. As for any illegal content, hosting providers must remove hate speech content promptly upon receiving notice, under the “notice and takedown” regime. Otherwise, victims may apply for blocking orders before courts. The court can then order internet access providers to block access to the relevant sites.
French audiovisual policy
34Both French and European law impose “must carry” obligations on telecom operators (a term that includes internet access providers) that distribute audiovisual programs.10 The operators must distribute certain public service television channels to all subscribers, generally free of charge. The must carry obligation is limited to certain qualifying television channels that serve a general interest.
35Audiovisual policy is also promoted via an obligation on providers of “on demand audiovisual media services” to invest a certain amount of their revenues in French and European production, to include in their catalog a majority of European works, and to present French and European works on the service’s home page. Under the country of origin rule established by the Audiovisual Media Services Directive11, these obligations only apply to providers of on-demand audiovisual media services established in France.
Apparent lack of coherence
36The foregoing summary of measures applied in France to advance content policies on the internet is a great simplification. The purpose of the summary is to illustrate the diversity of mechanisms currently used within a single country, and the lack of any apparent methodology to explain the differences in approaches. An invisible methodology may be at work. Each measure adapted by regulators takes into account prior experience, constitutional constraints, political realities, and international benchmarks. But there’s no visible roadmap. This leads to questions: Why is an independent regulatory authority used for some content (e.g. illegal gambling) but not for other content (e.g. hate speech)? Why is a court order required for some forms of blocking (e.g. copyright infringement), but not for others (DNS blocking for child pornography, or CNIL right to be forgotten orders)? Where has self-regulation been the most successful? Why are internet access providers targeted for certain kinds of actions, and search engines targeted for others?
37There are no doubt good reasons why different solutions apply to different content problems. However, policymakers approach each problem in isolation and can give the impression of reinventing the wheel each time. Without a baseline methodology against which to measure regulatory proposals, the solutions appear inconsistent and uncoordinated. Moreover, the measures cannot easily be compared to gather knowledge on what works, and what doesn’t.
Institutional authorities | Principal internet intermediary | Actions | Self regulation? | |
| Copyright infringement | HADOPI | Internet access provider (IAP) | IAP provides user identification. HADOPI sends warning notices. | Advertising service providers |
| Courts | Hosting providers, search engines, IAPs | Removing and/or blocking content | ||
| Right to be forgotten | CNIL, without court order | Search engines only | Delisting certain search results | Yes – internal search engine procedures to apply CJEU Costeja decision |
| Illegal gambling | ARJEL with court order | IAPs | DNS blocking | |
| Child porn and terrorism | Police without court order | Hosting providers, IAPs, search engines | DNS blocking by IAPs | Yes (INHOPE) |
| Hate speech | Courts | Hosting providers, IAPs, search engines | Yes | |
| Audiovisual policy | CSA | Cable operators, on-demand AVMS providers | Must carry.AVMS measures to promote European content |
Diminishing role of broadcasting regulation
38Another factor is at work, which is the diminishing role of audiovisual regulation in advancing national content policies. Historically, the licensing of broadcasting spectrum was the best way to ensure compliance with a wide range of national content policies. In addition to prohibiting illegal content, broadcasting licenses include rules to promote media diversity, plurality of opinions, to protect national security (via foreign ownership limitations), minors, public health, culture, language and national cinema. In some cases, broadcasting rules are designed to protect other economic sectors. The range of content policies contained in broadcasting licenses goes from matters of great national importance, such as rules protecting the proper functioning of democracy, to matters involving narrow economic interests, such as the protection of advertising revenues for regional newspapers. For decades, the broadcasting license has been a convenient basket in which politicians could throw numerous content rules designed to satisfy various stakeholders.
39The license to use spectrum is a convenient tool. Spectrum is part of the public domain. It belongs to the government, so the government can legitimately impose conditions in connection with its use. Just as the government can impose building rules for beachfront property designed to protect cultural and environmental aesthetics, it can impose usage rules on spectrum designed to promote French culture. In France at least, broadcasting spectrum is licensed free of charge, whereas mobile broadband spectrum is licensed in exchange for a hefty license fee. The government imposes content rules on broadcasters in lieu of a license fee.
40Over-the-air broadcasting still commands a large share of audience and viewing time in France. But its influence is diminishing, and will one day disappear. Some day in the future, most viewers will consume content online, using connected TVs, smartphones or tablets. Over-the-air channels received via a rooftop antenna will become the exception. Most broadcasters will make their content available via broadband (fiber, DSL, 4G/5G). When content providers no longer need broadcasting spectrum, governments will no longer have an easy “hook” through which to impose content policies. Governments will have to look elsewhere, and will turn to telecom operators and other internet intermediaries to fill the regulatory void. In 2006 Eli Noam predicted that telecommunication regulation will “become” broadcasting regulation, and that telecom operators will be asked to enforce national content policies because they are the only entities that regulators can reach within their jurisdictions (Noam, 2006). In fact, Noam’s prophecy can be applied not just to telecom operators, but to any internet intermediary that falls within a regulator’s jurisdictional reach. The decline of broadcasting regulation as a tool to regulate access to content explains why regulators look increasingly to internet intermediaries for solutions.
A methodology against which to measure regulatory proposals
41As the examples above show, lawmakers adopt measures to address a particular problem. Sunstein (1996) refers to this as the “pollutant of the month” syndrome. The measures create controversy and are often challenged in court. The measures are often accused of disproportionately harming fundamental rights, being costly and ineffective, and/or threatening the internet ecosystem (Haber, 2010). There lacks today a theoretical benchmark against which to test the relevant measures – whether imposed by government or voluntary – to ensure that they are as efficient as possible, and harm fundamental rights as little as possible.
Technical measures are inevitable
42We will start from the premise that national measures to involve internet intermediaries in the enforcement of content policies are inevitable (Noam, 2006). The question is not whether they will emerge, but how they should be built (Mann & Belzley, 2005).
Technical measures create harmful side-effects
43As noted above, measures implemented by internet intermediaries to block or limit access to illegal content can create negative externalities, including adverse effects for fundamental rights, net neutrality, and the internet ecosystem. Like a potent medicine, measures affecting internet intermediaries must be prescribed with care in order to avoid dangerous side effects. A blocking measure targeting illegal content can also block legal content. Some technical measures create privacy risks for users, and/or significant costs for internet intermediaries. Some measures may simply be ineffective. A government-imposed measure may also set a precedent for other countries, thereby creating an international arms race in regulation that could threaten or destroy the open character of the internet.
A reference methodology will help avoid mistakes
44If we accept as a given that certain technical measures are necessary, we need to consider what form should those measures take, and who should be in charge of administering them so as to minimize their harmful side effects.
45I propose in this book a reference methodology that could be used to evaluate regulatory proposals that affect internet intermediaries. The methodology would involve a questionnaire designed to help policymakers define the problem to be addressed, the alternatives available to address the problem and the direct and indirect costs generated by each alternative. The questionnaire would be followed by a cost benefit analysis (CBA), and the application of certain constraints. The questionnaire and cost-benefit analysis would then be subject to public consultation and to an institutional peer-review process. Finally, the regulatory measure would be subject to periodic ex post reviews to ensure that the measure is delivering its expected benefits and is not creating unexpected side-effects.
Existing literature
46The proposal builds on five categories of existing literature:
- Law and economics literature dealing with the most efficient level of law enforcement (Shavell, 1993) and most efficient level of ISP involvement to fight copyright infringement and other forms of illegal content (Lichtman & Landes, 2003; Lichtman & Posner, 2006; Mann & Belzley, 2005);
- Law and economics literature on net neutrality (Wu, 2003; Yoo, 2005; Curien & Maxwell, 2011);
- Law and economics literature on better regulation, the new public management and cost benefit analyses, with particular emphasis on environmental, health and safety regulation (Breyer, 1982; Sunstein, 1996; Hahn, 2004; Hahn & Litan, 2005; Posner, 2002; Ogus, 1998; Hancher, Larouche & Lavrijssen, 2003; Renda et al., 2013);
- Literature on institutional alternatives for internet governance, including self-regulatory and co-regulatory structures (Marsden, 2011; Brousseau, 2007; Weiser, 2009; OECD, 2011);
- Literature on the principle of proportionality, and the balancing of fundamental rights (Hickman, 2008; Tranberg, 2011; Sauter, 2013; Monaghan, 1970; Lemley & Volokh, 1998; Callanan et al., 2009).
How the remaining chapters are divided
47The remainder of this book is organized as follows:
48Chapter 2 will make an inventory of the factors that must be weighed when developing regulatory proposals. Chapter 2 will list the national content policies for which regulators may be tempted to enact regulation. Enforcement of those content policies is the expected benefit of regulation. Chapter 2 will establish a list of technical intermediaries and technical measures that can potentially be used to implement content policies. These are the technical tools that regulators might consider using. Chapter 2 will mention institutional alternatives and list the negative side effects that government-imposed technical measures might cause. These are the potential costs of regulation.
49Chapter 3 will focus on the balancing of fundamental rights. The European Court of Justice, the European Court of Human Rights and the United States Supreme Court afford to internet users a high level of freedom of expression. Internet content is entitled to the highest level of protection, similar to protection afforded to print media. The imposition of broadcast-style regulation on internet content would likely be illegal, violating the European Convention on Human Rights.
50Other fundamental rights in the balance include:
- the right to protection of property, which can justify measures designed to fight online copyright infringement;
- the right to security, which can justify measures designed to fight child pornography and terrorism;
- the right to privacy, which can be infringed by invasive technical measures, but can also justify measures taken to enforce national privacy laws (e.g. the right to be forgotten);
- the right to freely conduct a business, which can be interfered with when burdensome technical measures are imposed on internet intermediaries, or regulators proscribe certain business models.
51Each technical measure can enhance or impair fundamental rights. Courts have developed a balancing test to weigh these rights against each other, and determine a measure’s acceptability under constitutional principles. Chapter 3 will explain the proportionality test used by European courts, and how the test has been applied to technical measures designed to limit access to illegal content. Chapter 3 will also examine how law and economics scholars approach the fundamental rights of privacy and freedom of expression, which are the two rights most directly affected by technical measures imposed on internet intermediaries.
52Chapter 4 will examine the institutional aspects of technical measures, including the four main categories of institutional frameworks that can be used for regulating access to internet content:
- General liability or property rules enforced by the courts, which I call “court regulation”;
- Detailed regulatory rules developed and enforced by an administrative or regulatory body, which I call “administrative regulation”;
- Self-regulatory regimes, which can involve unilateral regulation by each firm through individual terms of use (“unilateral self-regulation”), and regulation through collective codes of conduct (“multilateral self-regulation”);
- Co-regulatory regimes, in which self-regulatory measures and government-imposed measures work together (“co-regulatory systems”).
53These four institutional frameworks often coexist with, and complement, each other. Indeed the first framework, general liability or property rules enforced by the courts, almost always exists, either by itself or as a backstop for other regulatory measures. In the shadow of liability rules and court enforcement, private actors use unilateral self-regulation, i.e. regulation through terms of use, to govern their relationship with users. The main additional options are administrative regulation, multilateral self-regulation and co-regulation. Chapter 4 will identify the advantages and disadvantages of these various institutional options, and how they can be used to deal with internet content issues. The objective is to put institutional options into the mix when evaluating regulatory alternatives. Brousseau’s (2007) work on multi-level governance will be discussed.
54Chapter 5 will focus on better regulation methodology, regulatory impact assessments and cost benefit analyses. Both the United States and Europe have developed methodology for testing regulatory measures using a cost-benefit analysis. The rigor with which the cost-benefit analyses are applied varies. Chapter 5 will examine the cost-benefit methodology and “good regulation” criteria proposed by the OECD, the United States government and the European Commission.
55A study by Renda et al. (2013) examines how cost benefit analyses should be conducted in the context of EU “better regulation” guidelines. Government Circular A4 describes the nuts and bolts of regulatory cost benefit analyses in the United States. Chapter 5 will also present the views of authors who argue that regulatory impact assessments are incompatible with the realities of the political process.
56Chapter 6 will attempt to bring together the themes of the preceding chapters by presenting a system for evaluating regulatory proposals affecting internet intermediaries. The system is inspired by the European Framework for regulating electronic communications and by the European Commission’s methodology for conducting regulatory impact assessments. The system consists of five parts:
- A questionnaire requiring policymakers to identify what the regulator hopes to achieve, how success will be measured, the regulatory options available, and the main costs and benefits associated with each option;
- A cost-benefit analysis requiring the preparation of a baseline scenario of no regulatory action and a comparison of the costs and benefits of various regulatory alternatives to the baseline scenario. The cost-benefit analysis will include a step where additional constraints can be imposed, which may lead to the elimination of certain regulatory proposals;
- The third part of the system consists of a public consultation, inviting stakeholder comments to the proposed regulatory impact assessment. The regulator would then revise the impact assessment to reflect stakeholder comments.
- The fourth part of the system consists of a peer review process pursuant to which the regulatory impact assessment is reviewed by an independent body. Robust peer review is a critical element of any “better regulation” strategy. Peer review of this kind exists already for cost-benefit analysis performed in the United States and regulatory impact assessments performed in the EU. In the field of electronic communications, a specific review system exists pursuant to which the European Commission reviews national regulatory proposals, and can in some cases veto them, prior to their adoption by national regulators. The system proposed in this book would require that proposed national measures affecting internet intermediaries undergo a similar review process.
- Finally, the system would require periodic ex post reviews to ensure that the measure continues to deliver the expected benefits in spite of technological and market developments. Regulations that are no longer fit for purpose would be amended or removed altogether.
57Chapter 7 will point out weaknesses of the proposed system and areas for future research.
58In conclusion, I will propose tools that will help determine what action (if any) is appropriate to deal with a given problem of harmful content, including what institutional framework, and what internet intermediary. The methodology will help trace lines, as illustrated on the by Figures 1 and 2 below.
59The methodology will necessarily be imperfect given the many variables involved and the difficulty measuring them in an objective manner. The methodology will also create extra costs of its own. Regulatory impact assessments, public consultations and institutional peer review take time and resources that could be invested elsewhere. These extra costs would have to be compared to the benefits derived from better rulemaking. Better rulemaking means laws and regulations that are effective, create fewer costs and adverse side effects, are more predictable, consistent, and future-proof.
60Applying cost-benefit analyses to internet regulation will help make trade-offs more visible, avoid costly errors, and will lead to the emergence of international best practices. Without the methodology, the accumulation of uncoordinated national measures will lead to the balkanisation of the internet.
Figure 1: A combination of institutional framework and technical measure to deal with hate speech

Figure 2: A combination of institutional framework and technical measure to deal with copyright infringement

Notes de bas de page
1 Directive 2002/21/EC.
2 CJEU Case n° C-131/12, Google Spain v. AEPD and Costeja, May 13, 2014.
3 French Law N° 78-17 of January 6, 1978.
4 Haute Autorité pour la diffusion des oeuvres et la protection de droits sur internet, created by Law N° 2009-669 of June 12, 2009.
5 A copy of the charter is available here: http://www.alpa.paris/wp-content/uploads/2001/01/20150323_MCC-signature-charte-publicite-1.pdf (consulted January 8, 2017).
6 Autorité de regulation des jeux en ligne, created by Law n° 2010-476 of May 12, 2010.
7 Decree n° 2011-2122 of December 30, 2011.
8 Law n° 2014-1353 of November 13, 2014.
9 Decree n° 2015-125 of February 5, 2014.
10 Article 34-1, Law n° 86-1067 of September 30, 1986; Article 31, Directive 2002/22/EC.
11 Directive 2010/13/EU.
Le texte seul est utilisable sous licence Licence OpenEdition Books. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.
Smart(er) Internet Regulation Through Cost-Benefit Analysis
Ce livre est diffusé en accès ouvert freemium. L’accès à la lecture en ligne est disponible. L’accès aux versions PDF et ePub est réservé aux bibliothèques l’ayant acquis. Vous pouvez vous connecter à votre bibliothèque à l’adresse suivante : https://freemium.openedition.org/oebooks
Si vous avez des questions, vous pouvez nous écrire à access[at]openedition.org
Smart(er) Internet Regulation Through Cost-Benefit Analysis
Vérifiez si votre bibliothèque a déjà acquis ce livre : authentifiez-vous à OpenEdition Freemium for Books.
Vous pouvez suggérer à votre bibliothèque d’acquérir un ou plusieurs livres publiés sur OpenEdition Books. N’hésitez pas à lui indiquer nos coordonnées : access[at]openedition.org
Vous pouvez également nous indiquer, à l’aide du formulaire suivant, les coordonnées de votre bibliothèque afin que nous la contactions pour lui suggérer l’achat de ce livre. Les champs suivis de (*) sont obligatoires.
Veuillez, s’il vous plaît, remplir tous les champs.
La syntaxe de l’email est incorrecte.
Référence numérique du chapitre
Format
Référence numérique du livre
Format
1 / 3