Access to evidence of discrimination has historically been seen as one of the greatest challenges for civil rights plaintiffs.
Gathering such evidence will likely become even more difficult as transactions that traditionally occurred in person—like renting homes, applying for loans, and buying goods—go online.
Discrimination has taken on a new face, hiding behind algorithms that crunch big data,
which is the data culled from individuals’ online activity.
And as discriminatory tactics evolve, so must the ways in which they are rooted out.
To that end, researchers have developed a number of methods to assess whether algorithms behave discriminatorily.
Many draw upon the conventional method for uncovering discriminatory practices: the use of testers, individuals who pose as potential consumers, renters, or employees for the sole purpose of testing whether an entity engages in discriminatory treatment in violation of civil rights statutes.
Two methods of testing—the “scraping” audit and the “sock puppet” audit—are particularly promising.
In a “scraping” audit, a researcher uses a bot to impersonate users of various backgrounds; the bot then issues repeated queries to see how an algorithm functions in response and subsequently collects the responses it receives.
A “sock puppet” audit essentially replicates a traditional testing scenario online: Researchers imitate users of various backgrounds with fake accounts or preprogrammed data profiles to test whether any differential treatment occurs.
These methods have already proven to be effective. Researchers investigating racial discrimination on Airbnb, an online platform that facilitates short-term rentals and homestays, conducted a study in which twenty user profiles were constructed and then used to send rental requests to approximately 6,400 Airbnb hosts.
The sole difference among the requests was the requesting user’s name.
Half the profiles used names that were more common among white individuals, while the other half used names more common among African American individuals.
Through this sock puppet audit, researchers discovered that the requesters with black-sounding names were 16% less likely than those with white-sounding names to have their requests accepted.
The discrimination was recorded among both cheaper listings and more expensive ones, and the majority of hosts who declined the users with names common among African American individuals had never hosted an African American guest, suggesting the discrimination was prevalent and entrenched, especially among certain hosts.
The issue with these methods is that the Computer Fraud and Abuse Act (CFAA),
a cybersecurity statute, effectively prohibits their use. Many websites employ terms of service that prohibit scraping
and the creation of false profiles,
and courts have interpreted the CFAA to penalize violations of a website’s terms of service.
Thus, the CFAA presents a substantial obstacle to researchers, journalists, and civil rights activists hoping to use online civil rights testing methods. A recent case, Sandvig v. Sessions, brought this problem to the forefront. The American Civil Liberties Union (ACLU) filed the case on behalf of researchers and journalists who would like to engage in these methods of online civil rights testing but are chilled from doing so as a result of the CFAA.
The case raises a novel theory: The CFAA is unconstitutional because the testing activity it effectively prohibits is protected by the First Amendment.
This Note assesses these freedom of speech and expression claims with regard to online civil rights testing and argues that, in light of past precedent related to protected conduct, false speech, investigative journalism, and the tradition of testers in civil rights enforcement, First Amendment protections should extend to civil rights testing. Therefore, applying the CFAA to online civil rights testers infringes upon testers’ First Amendment rights. The focus of this Note is limited to the substantive question of whether civil rights testing, and more specifically online testing, is a type of conduct that the First Amendment reaches; it does not dwell upon related issues that may arise over the course of litigation.
Part I describes specific instances of algorithm-based discrimination that have been discovered and outlines the history of civil rights testing and First Amendment doctrine to understand the CFAA–First Amendment problem. Part II analyzes the act of online civil rights testing under the doctrinal frameworks for false speech and conduct incidental to speech, each of which could provide a basis for First Amendment protection of testing conduct; it concludes that the CFAA, as applied to testers, and the First Amendment are in tension because online civil rights testing can be protected under both frameworks. Finally, Part III argues that First Amendment protections should be pursued through litigation even though the CFAA could be amended to allow for a tester exception. Whether the CFAA is amended, basing citizens’ right to test for online discrimination in constitutional law will offer civil rights advocates, journalists, and researchers firmer, more lasting protection in accessing information about, and working against, discrimination in the age of big data.
I. Algorithm-based Discrimination, Testers, and the First Amendment
This Part contextualizes the filing of Sandvig v. Sessions. Namely, it highlights the emerging and widespread issue of algorithm-based discrimination, the history and importance of civil rights testing, and the doctrinal background that will inform an assessment of whether First Amendment protection covers the conduct of civil rights testing. Section I.A.1 discusses the recent phenomenon of online algorithms and their discriminatory impact to date. Section I.A.2 turns to civil rights testers, describing their traditional function and how testing has been retooled to address algorithm-based discrimination. Section I.B then summarizes the legal frameworks for two different lines of First Amendment case law, each of which could protect these new, online methods of civil rights testing.
A. Algorithm-based Discrimination and the Tool of Testers
1. Big Data and the Advent of Online Algorithm-based Discrimination. — Though the concept of big data is relatively new, it has become prominent in discussions of the internet and technology, especially in relation to online privacy. As used in this Note, big data is the “massive quantities of information produced by and about people, things, and their interactions,” culled from the “digital traces left by people” through their online activities.
As these data become increasingly available, tools for leveraging them develop as well.
Algorithms are one such type of tool. A number of companies and groups are turning to algorithms to synthesize large amounts of data into valuable predictions.
This can often be useful to consumers: For example, algorithms can be used to predict what books a consumer may like based on her past responses to suggestions.
However, algorithms can also be keyed to search for certain characteristics and thereby enable discrimination.
Algorithm-based discrimination, as used in this Note, is defined as the unequal results generated by the explicit or implicit biases built into algorithms. Explicit bias, or disparate treatment, occurs when algorithms are created or used to purposely discriminate against individuals. For instance, one linguist has developed software that can both predict an individual’s ethnicity based on the individual’s full name, address, and ZIP code and identify 158 ethnicities based on predictive algorithms that use different ethnic groups’ name patterns and demographic data.
Though this software’s contracts specify permissible uses to prevent any possible “suspicious motivations,”
not every company that develops or uses predictive algorithms may safeguard against discriminatory usage.
In the absence of this kind of disparate treatment, or purposeful discrimination, algorithms can nonetheless lead to discrimination by causing a disparate impact against certain groups.
For instance, the use of proxies can result in discrimination when certain information corresponds with membership within a particular protected class.
Disparate impact can also occur when an algorithm contains implicit bias, such as when it is based on examples
in which prejudice played a role in the decisionmaking.
While disparate impact is not always legally actionable,
testers who uncover disparate impact can still use their findings to seek change through nonlegal avenues.
Researchers have found or suspected that algorithms are working discriminatorily in a number of contexts. For instance, advertisers using algorithms may target groups in prejudicial ways by presenting different advertisements to different individuals based on their sex or race.
This was demonstrated in two recent examples: Research showed that Google displayed an advertisement for high-income jobs more often to men than to women,
and advertisements for arrest records were significantly more likely to show up on searches for historically black fraternities or for names considered “distinctively black.”
Algorithms may also lead to discriminatory outcomes by differentiating the prices of goods
based on protected characteristics like race or ethnicity.
For example, The Princeton Review, a test-preparation-services company, used a differential pricing model that showed different prices to individuals across various geographic areas.
This model, however, mapped onto racial demographics: Customers in geographic areas with higher Asian populations were charged more for online SAT tutoring.
This pricing scheme was likely not based on income, since lower-income neighborhoods with large Asian populations were still shown higher prices.
The ProPublica article that initially broke this story referred to the price differential as a “tiger mom tax”
—referencing a familiar stereotype of Asian American parents
and raising the uncomfortable idea that The Princeton Review sought to profit based on this stereotype.
and housing are additional areas in which algorithms can lead to discriminatory effects. In the recruiting context, algorithms could cause disparate impact if they are keyed to look for certain characteristics, such as academic credentials, and the schools they are programmed to select for have a stark lack of diversity.
Alternatively, algorithm-based “talent-matching” tools could cause purposeful discrimination if the algorithms they use to match employees with employers take into account employers’ prejudicial preferences.
Similarly, since big data is also used to facilitate housing transactions,
these data, which could come from both public sources like the Census Bureau and smartphone apps that draw data from users’ recorded locations, can lead to purposeful discrimination by “racially steer[ing] consumers—to rent or buy in Black neighborhoods or to stay away from them.”
The financial sector may also engage in algorithm-based discrimination when assessing applications for lending services.
For instance, new software for bankers uses big data to interpret signals about behavior (such as whether customers use proper capitalization) to aid in assessing creditworthiness.
Such software could discriminate against certain racial or ethnic groups, even if not programmed to do so, if a particular behavior that is taken into account is more prevalent among certain protected classes.
The above instances show the potential for big data to be used in perpetuating civil rights violations.
Unanticipated uses of big data, especially as these online tools become more sophisticated, amplify the threat of discrimination, and plaintiffs have begun to bring antidiscrimination suits as a result.
In order to monitor this new form of civil rights violations, advocates will have to update a traditional tactic for rooting out discrimination: testers.
2. The Tester Tradition. — The use of testers, individuals whose only aim is to test whether a particular entity is engaging in unlawful discrimination by posing as a potential patron, employee, or tenant,
has become a vital tool in detecting discrimination in areas such as housing
The Supreme Court recognized the role of testers in civil rights suits in Havens Realty Corp. v. Coleman,
holding that testers have standing to bring claims under section 804(d) of the Fair Housing Act.
Following Havens Realty, a number of courts of appeals recognized tester standing under other civil rights statutes.
In doing so, these courts highlighted the importance of testers in achieving the antidiscrimination ends Congress intended with these statutes. The Seventh Circuit, for instance, noted that testers function akin to private attorneys general, and “because proof of discrimination is often quite difficult to muster . . . testers provide evidence that, we have recognized, ‘is frequently valuable, if not indispensable.’”
Similarly, the Ninth Circuit acknowledged that “[t]esters have played a long and important role in fair housing enforcement.”
As these appellate courts have firmly recognized, testers’ importance comes from their truth-revealing power—that is, their ability to uncover discriminatory practices not readily apparent.
For the task of uncovering discrimination in the virtual realm, researchers have adopted and adapted the tool of testers. The two methods that form the focus of this Note are the “scraping” audit
and the “sock puppet” audit,
both of which can involve impersonating an authentic online patron through the falsification of big data, whether done by researcher-created bots or the researchers themselves. While these methods of testing have been considered the most fruitful for rooting out discrimination,
researchers have been hesitant to fully endorse the practices due to the threat of violating the Computer Fraud and Abuse Act.
As will be discussed in Part II, the CFAA prohibits any act in contravention of a website’s terms of service, and many websites have banned the sock puppet and scraping methods in their terms.
In order to show how this interpretation of the CFAA conflicts with the First Amendment, this Note next turns to a review of the First Amendment doctrine.
B. First Amendment Doctrine
The First Amendment protects, among other things, “freedom of speech, or of the press,”
liberties considered to be of the most precious and fundamental nature in the United States.
While the fifteen words that create the First Amendment’s speech and press protections seem relatively straightforward, First Amendment doctrine is in a confusing and unclear state, partly because there can be many conflicting rules within the case law.
Additionally, many activities protected under the First Amendment are not explicitly written into the Constitution.
With this in mind, this section will delineate some of the relevant strains of First Amendment jurisprudence and set forth the various frameworks used. Because civil rights testing is conduct, rather than a purely verbal act, this section will look to First Amendment protections for conduct incidental to speech.
However, since a case can be made that testing also involves a verbal act in the form of misrepresentation, this section will discuss false speech jurisprudence as well.
2. Conduct Incidental to Free Speech. — One line of First Amendment jurisprudence focuses on conduct incidental to, or in preparation for, speech.
This type of conduct has also been called newsgathering or information gathering in some cases,
because journalists and activists in these cases seek access to information for the purpose of subsequently engaging in speech. Under the doctrinal strand of conduct incidental to speech, the Court has protected the means of various kinds of speech, whether those are monetary contributions to political campaigns
or access to trials.
One aspect of the doctrine is focused on the right of access to governmental affairs. This line of cases sprang from Richmond Newspapers, Inc. v. Virginia, a case that found a First Amendment right to attend criminal trials.
In coming to its decision, the majority focused on the necessity of access in order to write about trials
and explained that the First Amendment should be interpreted broadly.
Subsequently, the Court created a two-prong test to determine whether a First Amendment “right of access” exists, looking first to whether the press or public historically had access to the process, and second to whether public access plays a significant role in the process’s function.
This test has also been applied by courts of appeals to governmental proceedings other than trials.
More generally, a right to gather information was referenced in Branzburg v. Hayes.
While the Court acknowledged the existence of such a right, it did not flesh out the contours of the right nor what it covered.
However, it is clear that there are limits: Earlier, the Court had stated that a right to information was not unlimited because “[t]here are few restrictions on action which could not be clothed by ingenious argument in the garb of decreased data flow.”
In determining where the line for protection can be drawn, Professors Alan Chen and Justin Marceau argue every action necessary to speech falls somewhere along a spectrum of activity.
On one end lie “the most basic elements of conduct that are necessary to engage in communication.”
Many actions on this end will not qualify for First Amendment protection because they will be “too attenuated from the actual expressive activity.”
An example of this would be the purchase of gasoline to drive to a rally.
On the other end lies “the directly communicative element of the expressive process”; an example of this would be the exhibition of videos or art.
Even if, however, conduct falls closer to the “directly communicative element” end of this spectrum and directly implicates the First Amendment, it may also implicate privacy concerns that are crystallized in tort and criminal laws.
Thus, courts have still held investigative reporters liable for invasion of privacy or other torts they committed in the course of newsgathering. For instance, in Dietemann v. Time, Inc., the Ninth Circuit ruled on a case that involved undercover Life Magazine reporters who posed as patients and took recordings and photographs during their medical visit in order to expose the plaintiff practicing medicine without a license.
In deciding whether the reporters violated California privacy law, the Ninth Circuit found that when one invites another into their home, they “take a risk that the visitor may not be what [they] seem” and that this visitor will retell what was heard and seen upon leaving.
However, one does not take the risk of being photographed or recorded, and the First Amendment does not immunize journalists from liability for this kind of violation.
Somewhat contrary to Dietemann, the Seventh Circuit, in Desnick v. American Broadcasting Cos., held that use of test patients, sent with concealed cameras to eye doctors’ offices to see whether improper diagnoses occurred, did not give rise to claims for trespass, invasions of privacy, or wiretapping under Illinois law.
The case distinguished Dietemann by claiming that here, the patients entered offices open to anyone seeking treatment and taped physicians engaging in professional, rather than personal, communications, whereas Dietemann involved a much more private context.
The Seventh Circuit also analogized to testers in the civil rights context in deciding that no trespass occurred, explaining that civil rights testers who enter property without disclosing their true intentions do not interfere with ownership or possession of land.
Courts of appeals that have addressed similar cases have agreed with Desnick,
though scholars disagree about whether information gathering should be protected.
As Professor Jane Bambauer notes, “To this day, the right to access information is underdeveloped.”
The case law indicates that courts believe newsgathering or information gathering is conduct incidental enough to speech that some constitutional protection must be granted, yet they are hesitant to develop the doctrine further. Under the spectrum theory that scholars have posited, the attenuation of the link between the conduct and the speech act is key.
However, even in cases in which the First Amendment is implicated, lower courts that have addressed the issue have focused on considerations of privacy and used tort law as an external limit to information gathering, though in practice this limit seems lenient.
2. False Speech and Misrepresentations. — The act of lying is another form of conduct that implicates speech.
The First Amendment protects some, but not all, types of lies.
The Supreme Court has rejected First Amendment challenges to fraud claims,
criminalization of one’s misrepresenting status as a government official,
and regulations about false or misleading commercial speech.
Traditionally, the First Amendment had not protected false statements if those statements were valueless.
Yet the Court’s decision in United States v. Alvarez makes clear that false speech must be more than merely valueless to be restricted—it must be fraudulent, defamatory, or cause some other “legally cognizable harm.”
With the move away from the traditional inquiry based on the speech’s value
and the Court’s “fractured” opinion, the exact analysis to apply in false speech cases is uncertain.
The Alvarez Court did stress, however, a desire to avoid chilling speech with governmental censorship
—unless such speech “was used to gain a material advantage.”
The Supreme Court has yet to address “the relevant constitutional implications of a common law misrepresentation action directed against a media defendant,”
so analyzing these questions requires recourse to lower court doctrine. When it comes to misrepresentations to gain access for an investigative purpose, the Ninth Circuit in Animal Legal Defense Fund v. Wasden relied on Alvarez to invalidate portions of Idaho’s “Ag-Gag law,” which made it a crime to, among other things, use misrepresentations to knowingly enter an agricultural facility, obtain records, obtain employment with the intent to cause economic or other injury, or make recordings via misrepresentation.
The court held that Idaho could not criminalize entry by misrepresentation, for entry onto property, even private property, is not a material gain.
In justifying its reasoning, the court referred to Food Lion and Desnick, cases in which courts held that testers and undercover journalists had not committed any tort violations.
In fact, the Desnick court explicitly stated there was no fraud in a tester scheme because “a scheme to expose publicly any bad practices that the investigative team discovered” was not fraudulent.
On the other hand, the Wasden court held that there was no First Amendment bar to Idaho’s criminalization of misrepresentations made to obtain records or misrepresentations made to obtain employment when the speaker intended to cause economic or other injury to the facility, because both records and employment are material gains.
The Ninth Circuit stated that while it may be true that undercover journalists seek to expose public threats rather than secure material gain, because the Supreme Court had specifically referred to offers for employment when giving examples of material gains and because undercover investigators are paid by the facilities as part of their employment, they are not immune from Alvarez’s holding.
The court stressed, however, that “this does not mean that every investigative reporter hired under false pretenses intends to harm the employer,” and therefore, so long as the requisite intent is not present, journalists would be protected.
Thus, though the Ninth Circuit only partially invalidated Idaho’s ag-gag law, it nonetheless expressed caution in allowing criminalization based on misrepresentations, evidenced by its strict reading of what constituted “material gains” and its emphasis on the intent requirement for criminalization of misrepresentations for employment.
Given this, while the Supreme Court has not addressed the specific issue of misrepresentation claims and newsgathering or civil rights testing, lower courts before and after Alvarez seem wary of validating tort claims under a misrepresentation theory.
After reviewing the issue of algorithm-based discrimination, civil rights testing historically and online, and relevant doctrinal strands of First Amendment case law, this Note next turns to the conflict between the First Amendment and the CFAA.
II. Tension Between the CFAA’s Prohibition on Online Civil Rights Testing and the First Amendment
In late June of 2016, the ACLU filed a complaint on behalf of academic researchers and First Look Media Works
challenging the constitutionality of the CFAA.
The plaintiff researchers in Sandvig v. Sessions are actively attempting to research whether housing and hiring websites are engaging in discriminatory conduct,
and First Look Media Works is a journalism platform hoping to conduct similar research on websites’ discriminatory use of data and algorithms.
The complaint alleges some possible First Amendment grounds on which the CFAA could be invalidated as a result of its prohibition on online testing and raises both facial and as-applied challenges.
Though the plaintiffs’ facial challenge was dismissed,
the issue remains that insofar as the CFAA restricts these plaintiffs’ rights to test for online discrimination, this law runs up against the First Amendment because “research, testing, or investigations” constitute protected speech or expression.
In response, the government filed a motion to dismiss, arguing that the conduct in this case was nonspeech activity not reached by the First Amendment.
Though scholars have commented on the possible friction between the CFAA and the First Amendment for some years now, the particular issue raised by the plaintiffs in Sandvig remains unanswered. Previous scholarship has largely revolved around the First Amendment implications of the CFAA’s proscription of “hacktivism,” an act that may be protected by the First Amendment.
Other scholarship argues that the CFAA is an overbroad law that inadvertently infringes upon the protected speech of independent commentators online.
Additionally, the CFAA may be in tension with the First Amendment because it allows website owners to control information use and thus essentially confers private rights in information to the site owners, thereby restricting speech.
Finally, the most recent scholarship involving this topic has argued in theoretical terms that the First Amendment should encompass “Empirical Liberty,” or a “scientific” right to experiment and research.
Still, the scholarship exploring the tensions between the First Amendment and the CFAA is generally fairly limited in scope. It has not addressed the specific question raised by Sandvig v. Sessions: whether the CFAA infringes on the First Amendment rights of testers by preventing online civil rights testing. The remainder of this Note turns to this issue and contends that the CFAA conflicts with the First Amendment when applied to advocates, journalists, and researchers seeking to conduct online civil rights testing because such testing is protected by the First Amendment.
This Part first examines the CFAA and highlights how courts’ interpretation of this statute has resulted in the CFAA’s effective prohibition against online civil rights testing. The following section then outlines the various doctrinal bases that litigators could use to seek First Amendment protection for online civil rights testing and assesses the viability of each one, ultimately arguing that the theories of false speech and conduct incidental to speech can give rise to a First Amendment right for testers. Therefore, insofar as the CFAA is applied against online civil rights testers, such application infringes upon the First Amendment.
A. The Computer Fraud and Abuse Act (CFAA)
Congress passed the Computer Fraud and Abuse Act
in 1984 to address the issue of cybersecurity.
The CFAA prohibits computer trespass by preventing parties from knowingly accessing a protected computer
either “without authorization” or by “exceed[ing] authorized access.”
The scope of the CFAA is still an open question, with courts “divided deeply over when access is authorized.”
Defendants have been prosecuted under the CFAA for acts ranging from gathering information from an employer’s computer for personal use in violation of a company rule
to collecting information from a website after the site had sent a cease-and-desist letter and blocked the defendant’s IP address.
Some courts have also held that violation of a website’s terms of service can trigger the CFAA,
and the Department of Justice has supported this position.
The rationale is that website terms of service are meant to limit users’ access, so users who violate those terms thereby exceed “authorized” access.
Thus, this reading gives effect to website terms of service that prohibit the “sock puppet” or “scraping” online civil rights testing methods. For instance, websites, including those that provide housing and employment services, such as Zillow.com
prohibit the use of bots, scraping, and the creation of false accounts.
These sites could be engaging in algorithm-based discrimination, yet absent an exception for civil rights testing, testers using either the sock puppet or scraping method to detect such discrimination violate the sites’ terms of service and could thus face liability under the CFAA.
To protect online testing, the plaintiffs in Sandvig v. Sessions seek to strike down the CFAA on First Amendment grounds.
This Note next argues that because the First Amendment can in fact protect online civil rights testing, the CFAA, under courts’ current interpretation, infringes on constitutionally protected activities.
B. The Incompatibility Between the First Amendment and a CFAA Testing Restriction
Whether the CFAA conflicts with the First Amendment because of its restriction on online civil rights testing is predicated on a few issues,
the most important of which is whether civil rights testing of this kind is even conduct that could be protected by the First Amendment.
This section will look at the various doctrinal lines that could offer protection for testing conduct and assess the strength of each one. Section II.B.1 evaluates whether the First Amendment’s protection for false speech conflicts with the testing prohibition at issue here, and section II.B.2 grounds a claim for First Amendment protection in the idea that civil rights testing can be seen as conduct incidental to speech.
1. False Speech. — First is the argument that the CFAA, when applied to online civil rights testing, violates the First Amendment because it imposes criminal penalties on misrepresentation, or false speech, which is protected by the First Amendment in certain situations. Though online testing is technically conduct, it nonetheless can be seen as an act of false speech. Just as in-person testers make affirmative misrepresentations about their identities or intentions, online testers do essentially the same: The testers (or the bots they program) paint themselves as individuals who they are not for the purpose of gaining access to certain sites.
Due to these parallels, it is sensible to apply the false speech doctrine when analyzing claims of testers under the First Amendment.
Alvarez held that the First Amendment prevents criminalization of falsehoods that are not fraudulent or cause no tangible harm,
and the CFAA’s prohibition on testing conflicts with this holding.
Testing activity can be seen as an act of false speech by way of making a misrepresentation, as testers misrepresent their identities and intentions. Yet courts have found that the kind of misrepresentation traditional testers engage in is not fraudulent or harmful.
Furthermore, recent applications of Alvarez at the lower court level in ag-gag cases, such as Animal Legal Defense Fund v. Wasden,
also lend support to the Sandvig plaintiffs. Just as these ag-gag laws criminalize misrepresenting oneself as a worker or potential employee for the purposes of undercover investigations,
the CFAA effectively criminalizes misrepresenting oneself as a potential customer, tenant, or job applicant even for the purposes of online civil rights testing.
In Wasden, the Ninth Circuit said that misrepresentations made for entry were not made for material gain.
Though the court said that misrepresentations made for employment were made for material gain, the court emphasized that the statute in this case only criminalizes misrepresentations made with the intent to cause harm to the facility and therefore was not overbroad.
Protecting the misrepresentations made in online civil rights testing does not run afoul of the Ninth Circuit’s decision. Though it could be argued that employment discrimination testers may seek to cause the discriminatory employers they test legal or economic harm, when testers initially seek employment opportunities, they typically are unsure of whether a particular employer is actually discriminatory.
Thus, like an investigative journalist, a tester in most cases will not apply for employment with the preconceived intent to cause harm, and therefore, testers would not be subject to the Idaho statute as the Ninth Circuit described it. The case law accordingly supports the inference that the First Amendment protects the kind of investigative activity the plaintiffs in Sandvig seek to undertake.
Even if the misrepresentation for employment purposes involved in ag-gag law cases is not wholly clear-cut based on Alvarez, misrepresentations made during civil rights testing should not be considered analogous to the type of misrepresentation for gain the Court has said would fall outside the protections of the First Amendment, because misrepresentations made by testers are not fraudulent.
A rule that misrepresentations made during civil rights testing are actionable as fraud would have a chilling effect on all kinds of testing and undermine the important purpose of testers, which has been repeatedly recognized by the courts.
To that end, some district courts have explicitly recognized that testers posing as potential consumers or applicants do not engage in misrepresentation.
The most significant drawback to grounding testers’ First Amendment rights in the false speech framework is the ambiguity presented by Animal Legal Defense Fund v. Wasden. Though the court seemed to suggest that any misrepresentations made for employment are unprotected, it also repeatedly emphasized the specific intent requirement built into the statute and rebutted the Animal Legal Defense Fund’s argument that the statute was overbroad by pointing to the intent requirement.
Given the court’s treatment, it remains somewhat unclear whether the First Amendment would allow a categorical ban on misrepresentations for employment purposes, regardless of the speaker’s intentions.
Were courts to allow prohibitions on false speech for employment purposes irrespective of whether the speaker had a truth-seeking intent, the result would cause some internal tension within First Amendment law: In Alvarez, the defendant engaged in speech for no valuable purpose and yet his speech merited protection,
whereas those who engage in misrepresentations to examine potential legal violations are attempting to discover the truth and thereby perform a public service but would not be protected. Given the First Amendment’s strong roots in protecting truth-seeking and encouraging public knowledge and debate,
allowing for unqualified prohibitions on misrepresentations made for employment seems antithetical to the amendment.
Even if blanket bans on misrepresentations made for offers of employment were to be upheld, online civil rights testers may still be exempt. Unlike undercover investigators who seek to research an entity or industry by working within it, testers merely aim to assess whether differential treatment occurs in the hiring process.
Therefore, testers aren’t really seeking offers of employment. In addition, the Wasden court focused on the fact that “undercover investigators are nonetheless paid by the agricultural production facility as part of their employment.”
But civil rights testers would not actually be paid, as they are not planning on working for the prospective employer. Thus, were a categorical prohibition on misrepresentations made for employment to be upheld, it’s nonetheless feasible to protect testers.
Further counterarguments against protecting civil rights testers based on false speech doctrine can also be adequately addressed. First, some may argue that the Stolen Valor Act of 2005, which was at issue in Alvarez, directly criminalizes false speech just as ag-gag laws do,
whereas the CFAA imposes no such facial proscription. However, as applied in the case of civil rights testers, the effect is the same: The courts’ current interpretation of the CFAA effectively prevents testers from engaging in false speech even if it does not result in fraud or material gain. Thus, though the CFAA is not a facial proscription on speech, it nonetheless runs afoul of the First Amendment when it is interpreted to block tester activity.
For these reasons, under the current framework of the false speech doctrine, the First Amendment likely could protect misrepresentations made by testers in the course of their conduct. This, in turn, creates tension between the CFAA and freedom of speech—an interpretation of the CFAA that renders online testing illegal and thereby penalizes false speech runs counter to Alvarez’s holding.
2. Conduct Incidental to Speech. — The second line of doctrine that could offer an avenue for First Amendment safeguarding is that of conduct incidental to speech. Within this case law, there are two different approaches that could be used. First, there is the right of access line of cases that follows Richmond Newspapers, Inc. v. Virginia.
Second, there is the more general doctrine of conduct incidental to speech.
Although there is no formula for analyzing what constitutes conduct incidental to speech, perhaps because the Court has been hesitant to flesh out this right,
this section will use the spectrum analysis discussed by Professors Chen and Marceau,
along with lower court decisions, to assess the claim that testing conduct could be seen as incidental enough to speech to garner First Amendment protection. It also discusses potential barriers and the strengths of the analysis under each approach. Though it seems unlikely that Richmond Newspapers and its progeny could support a First Amendment right to conduct civil rights testing, the right can be firmly grounded in the doctrine protecting conduct incidental to speech.
a. First Amendment Right of Access. — While the argument for tester protection based on the line of right of access cases following Richmond Newspapers is plausible, it is unlikely to succeed. Such an argument would be grounded in the two considerations set forth by the Supreme Court: history and the role of public access.
Historically, civil rights testing occurred in business operations that not only were open to the general public but actually relied on the public coming in and out.
These businesses are not public areas in the sense that they are publicly or governmentally owned, but they have been historically open to public access.
With regard to the second prong, it could be argued that civil rights testing plays a significant role in the functioning of businesses and commerce.
In fact, passage of Title II of the Civil Rights Act, which prohibits discrimination in public accommodations, evidences the economic benefits of antidiscrimination, though the statute can be justified on both moral and economic grounds.
Yet a court would be unlikely to accept this argument because courts have not yet applied this test to extragovernmental contexts.
Richmond Newspapers focused on the historical, rather than functional, reasons for allowing a right of access, and the history largely revolved around the First Amendment’s “common core purpose of assuring freedom of communication on matters relating to the functioning of government.”
If the First Amendment purpose being served is the ability to use speech as a check on government actions, limiting the progeny of Richmond Newspapers to governmental affairs makes sense.
The Supreme Court recently decided a case involving access and the First Amendment—but rather than dealing with access to a proceeding or physical place, the case dealt with a sex offender’s right to access the internet.
Though the case primarily addressed the interplay between the internet and the public forum doctrine,
the Court nonetheless seemed to ground the right to access in the speech acts that the access facilitated.
The Court’s decision thus suggests that the doctrinal line involving conduct incidental to speech is a stronger basis for protecting civil rights testing, as discussed next.
b. Spectrum Analysis. — A more viable claim is that civil rights testing is protected as a kind of newsgathering or information-gathering conduct that falls close enough to speech to merit protection, even if there is no “unrestrained right” to gather information.
Due to the lack of guidance from the Supreme Court on when newsgathering actions merit protection,
this discussion will instead use lower court opinions and the spectrum theory posited by Professors Chen and Marceau.
Under the spectrum theory, conduct that falls closer to a final speech act is more likely to be protected.
With regard to civil rights testing, the level of attenuation between the speech activity (the publication of the testing results) and the conduct seeking protection (the testing itself) is insubstantial. In the context of an exposé, what is more directly antecedent to its writing and subsequent publication than the investigation that yielded its content? If campaign financing is seen as necessary enough to political speech to merit First Amendment protection, it makes little sense to say that civil rights testing is unnecessary for an investigative news or research article to be published, since it is this very testing that provides the content of such an article. Further, as Professor Bambauer notes, a rule that data are speech while their creation is not, when applied to other contexts, “would immediately draw suspicion: Selling books is protected speech; could printing them possibly be regulable conduct? Music is speech, but could it be that recording and producing an album is conduct?”
This theory has most recently reemerged in the form of “right to record” cases at the appellate level. Every circuit that has considered whether the First Amendment protects the right of individuals to record official police activity has answered in the affirmative.
These cases have rejected the artificial distinction between the videos or recordings themselves, which are unquestionably protected by the First Amendment,
and the act of making the recordings or videos. As the Third Circuit said, in order to give actual photos, videos, and recordings meaningful protection, “the Amendment must also protect the act of creating that material.”
The court went on to explicitly recognize that there is no practical distinction between preventing citizens from making a recording and banning the possession or distribution of recordings.
Other “right to record” cases have echoed this understanding that both the final speech act and its creation should be equally protected.
Thus, courts have seen prohibitions on recording as “necessarily limit[ing] the information that might later be published or broadcast—whether to the general public or to a single family member or friend—and thus burden[ing] First Amendment rights.”
Another weighty consideration pushing these courts to recognize a “right to record” was the notion that such recordings may assist in unveiling and correcting police misconduct. As one court noted, civilian videos have filled the gaps when recordings of police conduct are otherwise unavailable or withheld, and these videos have been critical in kick-starting reform and assisting civil rights investigations.
Even absent publication of the videos, the act of recording can improve policing.
Under a similar logic, lower courts have gone one step further and recognized that investigations are incidental to speech. One court found that a plaintiff’s evidence-gathering activities, which involved monitoring and photographing the city’s mayor, were protected conduct because they gave rise to her speech concerning the mayor and police department’s actions.
Similarly, a California appellate court found that hiring an investigator could be protected as conduct that “intrinsically facilitates exercise of free speech.”
Another court extended First Amendment protection to the conduct of interviewing a prisoner as part of the plaintiff’s project of writing the biography of James Earl Ray, holding a Bureau of Prisons ban on interviews to be unconstitutional.
Further, a court held that a statute prohibiting physicians from advertising their costs or fees, indirectly or directly, abridged the plaintiffs’ right to gather, receive, and publish such information.
Like in the “right to record” cases, these cases that more specifically deal with newsgathering similarly focus on the facilitative role of the investigative activities in question. These courts emphasize the “light [that] may be shed on” the information or ideas that will ultimately be published or broadcast
, 369 F. Supp. at 1186.
and the “in furtherance of speech” element inherent in newsgathering.
These courts have also been careful to take note of the public interest in the final speech act.
These elements on which the “right to record” and newsgathering cases turn can also be found in the context of online civil rights testing. As highlighted above, testing activity, though conduct, is necessary for the documentation of the results, which is the final speech act; just as recording video or audio is a means to the final film or tape, online civil rights testing is the means by which the final speech act is produced. Regardless of whether the results are published, the means by which these results are procured should therefore receive First Amendment protection.
If prohibiting the publication of civil rights testing results would infringe on the First Amendment, then so too does a prohibition on the testing itself. To hold otherwise would create an end run around the Constitution.
Additionally, online civil rights testing generates significant public value. Investigations in general have played a crucial role in exposing the truth in certain industries and practices, dating as far back as the early 1900s with Upton Sinclair’s The Jungle
and the work of other muckrakers.
Civil rights testing more specifically, however, serves a public function in the same way that recording the police can: Whereas police recordings can uncover police misconduct, online civil rights testing can uncover violations of civil rights statutes.
As it currently stands, online civil rights testing methods such as sock puppet and scraping audits are the most viable methods of checking whether algorithms comply with antidiscrimination laws.
Consequently, prohibiting this testing severely undermines the ability to monitor companies transacting online. And without the safe harbor from testing that the CFAA affords these companies, they may become more conscious of the disparate effects of their technology and ameliorate these effects voluntarily, just as the act of recording police officers can itself mitigate against misconduct.
Moreover, publicizing the issues surrounding algorithms and discrimination can also lead to public discourse on matters such as the evolving standards of civil rights enforcement or the drawbacks of current technology more broadly.
Thus, recognizing that online civil rights testing is protected conduct incidental to speech is squarely in line with the analysis of courts upholding a “right to record” or a right to newsgather.
There are some features of these “right to record” and newsgathering cases that may make them appear distinguishable from the context of online civil rights testing. Yet more recent cases suggest that these distinctions are immaterial. For instance, one could argue that recording audio or video in preparation for broadcast is more analogous to writing in preparation for publishing; in other words, data collection seems to be more degrees removed than recording. Though perhaps technically true, courts have nonetheless begun treating data collection as protected conduct,
suggesting that the difference between data collection and recording is inconsequential.
Second, these cases often involve public officials and public areas,
while online civil rights testing takes place in a private domain. But courts reviewing recently enacted ag-gag laws nevertheless have struck down prohibitions on recording on private agricultural facilities,
indicating that this too is an insignificant distinction. As one of these courts pointed out, the idea that speech acts on private facilities enjoy no First Amendment protection finds “no support in the case law.”
Rather, such an argument “confuses two related but distinct concepts: a landowner’s ability to exclude from her property someone who wishes to speak, and the government’s ability to jail the person for that speech.”
Thus, Utah’s criminalization of speech acts even on private property violated the First Amendment.
The First Amendment issues in the CFAA context are analogous. Though online civil rights testing takes place on private websites, it is criminalized by the CFAA: By placing criminal sanctions on conduct that violates websites’ terms of service, which in turn prohibit scraping and the creation of false profiles, the government punishes testers for engaging in speech acts in violation of the First Amendment. The constitutional issue turns not on the location of the speech but rather on the government’s sanctioning of it.
Protecting online civil rights testing conduct based on the theory that it is conduct incidental to speech would align with courts’ analyses of why recording and other types of newsgathering fall under the First Amendment’s purview. Not only is it immediately necessary for the final speech act, but civil rights testing, whether online or in person, also falls in line with one of the core purposes of the First Amendment: revealing the truth.
Therefore, the interpretation of the CFAA currently used by courts, which prohibits testing activity, infringes upon protected-speech-enabling conduct.
c. Additional Barriers. — While it is likely that testing activity could be seen as conduct incidental to speech, there are further challenges to extending First Amendment protection here that should be addressed. First, as the Dietemann case illustrates, the First Amendment is not a defense to torts or crimes committed by speech actors in the course of newsgathering.
Second, if testing activity occurs on private property, the state action requirement or public forum doctrine may affect the viability of a First Amendment claim. Third, because the scraping audit involves bots engaging in the actual data collection, one might argue that the First Amendment shouldn’t extend to scraping because bots’ activity can’t be constitutionally protected. Finally, extending protections to testers could raise issues in defining the scope of the right. However, all of these concerns can be adequately addressed.
While courts will likely not grant civil rights testing First Amendment protection when there are tort violations (so as not to extend newsgathering privileges to illegal conduct),
this concern is not as weighty as it seems. For one, as the Seventh Circuit wrote in Desnick, testers would not be subject to trespass claims because they do not interfere with ownership or possession in a way that trespass law intends to protect against.
The Ninth Circuit and Fourth Circuit have both commended this view,
and other district and state courts have also suggested or adopted this idea.
In addition, in practice, courts have been lenient with the press in the investigative journalism context, either by not recognizing that a right has been violated
or by giving next to nothing in the form of damages.
Part of this may relate to the Supreme Court’s constant concern that extensive tort liability will chill speech of constitutional value.
Furthermore, in the most recent appellate case addressing the interplay between investigators and torts, the Ninth Circuit in Wasden struck down a provision criminalizing misrepresentations made to gain entry to agricultural facilities on First Amendment grounds, citing Desnick and Food Lion but making no mention of Dietemann.
Given that testing involves accessing public, rather than private, domains, it seems more likely that a court would follow the Wasden court and rely on Desnick and Food Lion instead of Dietemann.
Second, with regards to the First Amendment’s state action requirement, the Supreme Court has noted that “[t]he test is not the form in which state power has been applied but, whatever the form, whether such power has in fact been exercised.”
Thus, even if a private litigant is seeking to enforce a tort at common law, for instance, a court’s application of that law can bring the First Amendment into play.
Therefore, that websites promulgate terms of service preventing the occurrence of conduct incidental to speech is not a shield from First Amendment scrutiny.
The public forum doctrine may also be concerned here. The public forum doctrine is invoked when speakers seek access to public property to engage in speech acts.
Thus, if websites are public forums, restrictions on the speech acts taking place on these websites, like online civil rights testing, would implicate the First Amendment. No case has explicitly reached the question of whether public-facing websites constitute public forums.
However, the Court’s recent decision in Packingham v. North Carolina
and the lower court rulings that have followed suggest that the public forum doctrine might encompass the internet and social media, an understanding that falls in line with what scholars have suggested.
The Packingham Court, in overturning a North Carolina statute that restricted sex offenders’ access to commercial social networking websites, noted that cyberspace has become the most important place for the exchange of ideas.
In likening social media websites to “the modern public square,” analogizing to streets or parks that were “quintessential forum[s]” for the exercise of First Amendment rights, and failing to mention the public–private nature of websites on the internet,
the Court implied that for First Amendment purposes, the internet might very well be considered a public forum. Lower courts since Packingham seem to have adopted this message,
suggesting that while the public forum doctrine might very well be implicated in the case of online civil rights testers, it may not pose an obstacle.
A third issue that might be raised is that in the scraping audit, the tester is not conducting the protected activity at issue; rather, bots are conducting the protected activity (that is, the investigation and subsequent data collection).
Nonetheless, a First Amendment right extends to the bots’ activities. Many lower courts have recently granted First Amendment protections for algorithmically produced search-engine results on the basis that the algorithms embody and perpetuate the speech of the engineers who create them.
Analogously, the bots that are created and directed by online civil rights testers should also receive First Amendment protections.
A final consideration to address is that of scope: How far could the arguments that give rise to First Amendment protections for civil rights testing reach? Could these arguments allow for protections for scraping or creation of false profiles to gather data that would be used for purposes other than civil rights enforcement? Given that many of the cases involving scraping and the CFAA deal with entities seeking confidential data from their competitors, companies may have significant concerns with the scope of the right advocated for in this Note. And under a plain reading of the case law, it’s unclear whether purpose or intent would be a notable factor in deciding whether the First Amendment reaches particular data-gathering conduct. The Supreme Court has explicitly dispelled the notion that the First Amendment specially privileges journalists,
and some of the “right to record” decisions have foundations in this very case law.
Going even further, some of these cases have explicitly disavowed the idea that the recorder must intend to disseminate the video at the time of recording.
At the same time, however, courts deciding these “right to record” and other cases addressing newsgathering conduct are at least partially motivated by the public interest served by the conduct at issue.
Thus, despite courts’ acknowledgment that the First Amendment protects journalists and nonjournalists alike, the fact that purpose has some analytical force suggests that scraping and creation of false profiles in violation of terms of service would not be protected in just any context. Though this context does not necessarily need to be enforcement of civil rights statutes, it may still need to have some significant public interest element, thereby cabining the scope of this right.
Whether testing activity is protected as conduct incidental to speech or as false speech, it nonetheless should fall under the First Amendment’s purview as a matter of doctrine and theory. Testing, like other investigative tactics, directly facilitates publication; yet unlike certain investigative methods, courts have not found civil rights testing to be a tortious act. Similarly, civil rights testing neither involves tortious misrepresentations nor gives rise to any legally cognizable harm such that testers’ false speech can be regulated.
Furthermore, the truth-seeking aims of civil rights testing align with the First Amendment’s purpose of ensuring an informed citizenry. When courts interpret the CFAA in a way that prohibits activity based on a website’s terms of service, which can include prohibitions on certain methods of online civil rights testing, the CFAA effectively restricts constitutionally protected speech activity and falls under the First Amendment’s scrutiny.
III. How to Protect Testers from the CFAA: Legislation or Litigation?
As established, civil rights testing is an enormously important tool for enforcing civil rights laws, and this conduct falls under the First Amendment’s protection. Thus, the current interpretation suggested by courts and the Department of Justice—that online civil rights testing is unlawful when prohibited by a website’s terms of service—may be unconstitutional. This Part analyzes possible solutions to this problem. Section III.A looks at the suggestion that Congress amend the CFAA to exclude civil rights testing, and section III.B addresses judicial resolution of the CFAA–First Amendment tension. This Part ultimately concludes that litigation is the optimal route to achieve the strongest protection of civil rights testing.
A. Legislatively Protecting Testers or Amending the CFAA
One solution to the problem of the CFAA’s prohibition on testing conduct is congressional legislation amending the CFAA so as to preclude courts from interpretations that proscribe civil rights testing. This appears to be an attractive solution because, as the CFAA is legislation itself, it would be simplest for Congress to overrule courts’ interpretations as it has done in the past.
Additionally, to protect civil rights testers more generally, Congress could use its Commerce Clause power to affirmatively shield testers from exposure to tort liability.
This would be an alternative to judicial recognition of a First Amendment right to engage in civil rights testing.
The main arguments in favor of using legislation rather than adjudication
center on the notion that there is a level of legitimacy conferred upon the decisions of a democratically elected majority that is absent in judge-made law.
Some scholars say that judges who adjudicate on matters involving social rights not grounded in the Constitution are considered to be “the closest thing the United States has to a governing nobility”
—antidemocratic actors whose decisions are indistinguishable from those of the Justices who decided Dred Scott and Lochner.
Thus, in the context of the CFAA, it could be argued that Congress is best suited to protect civil rights testers as a matter of legitimacy, for courts should not “find” new rights in the Constitution not readily apparent from the text or past precedent.
There are certain weaknesses in the argument that democratic political means should be the primary vehicle for social change. For one, the above criticism rests on the assumption that there is no textual basis for the right at stake.
In the case at hand, however, as the analysis from section II.B establishes, a First Amendment right to test for discrimination can be predicated upon an ample supply of doctrinal foundations. In other words, a tester’s rights can be firmly grounded in the Constitution’s text and past precedents. Furthermore, on a more conceptual level, although interpreting open-ended constitutional provisions can seem like legislating from the bench because there is little text to guide judges, this is not necessarily Lochnerism.
Another issue with the political process argument is that it seems to ignore the current political reality of extreme polarization and the resulting barriers to creating substantive legislative change.
Constitutional law scholars have argued that although the judiciary espouses a general skepticism toward innovation used to circumvent this polarization, such skepticism has no constitutional basis, and judges should take into account political reality when adjudicating constitutional cases.
Thus, while it intuitively seems that Congress is best suited for protecting civil rights testers, given the political reality of modern times, it may be that the courts are in fact the most reliable source of protection. Though barriers exist to a successful First Amendment challenge to protect testing via litigation, there are benefits to doing so that are not available with legislative remedies.
B. Litigating for First Amendment Protection
As the analysis in section II.B highlights, there are multiple lines of case law that can support First Amendment protection for civil rights testers. Thus, courts could interpret the CFAA in a way that avoids implicating the First Amendment altogether pursuant to the canon of constitutional avoidance, a canon which essentially states that courts should avoid ruling on constitutional issues when there is a way to resolve the issue on a nonconstitutional basis.
Currently courts use website terms of service to determine when an individual has engaged in “unauthorized” access because they consider a violation of terms of service to be the proper metric for finding a CFAA violation.
Courts could instead attempt to redefine such an interpretation when applied to testers, perhaps by analogizing to courts’ discussion of testers and torts
because the CFAA was meant to apply to the online equivalent of “trespass.”
Because courts have dismissed the notion that civil rights testers can incur tort liability for going onto private property and misrepresenting themselves as potential patrons,
courts analogously could dismiss actions against testers who violate websites’ terms of service (and thereby commit computer trespass) to conduct testing. Under the canon of constitutional avoidance, because courts could interpret the CFAA in a way that does not implicate the First Amendment, they should.
Were an opportunity to establish affirmative constitutional protection for civil rights testers available, however, it should be pursued.
Though the viability of arguments in favor of tester protection may be tempered by the relative incoherence of First Amendment doctrine,
the novelty of the arguments themselves, and the Supreme Court’s fairly deferential application of strict scrutiny in at least some First Amendment cases,
there are appreciable benefits to mounting a challenge to the CFAA on First Amendment grounds, so long as it is possible.
For example, there is a certain sense of permanence that comes from constitutional interpretation in that only the Supreme Court is able to reinterpret the substantive rights found in the Constitution
—absent an amendment, which is notoriously difficult to pass.
In addition, though a statutorily created right has a sense of legitimacy in that it derives from the authority of a democratically elected Congress, a constitutional right tends to be perceived, whether accurately or not, as even more legitimate and respectable because of the extremely venerable position the Constitution holds.
While the desire to use the legislature to enact social change is compelling, litigating for First Amendment protection would be the best strategy to protect civil rights testers as there are far more challenges to amending the CFAA or separately creating an affirmative statutory protection for testers. Litigation allows for a sense of permanence and legitimacy that does not necessarily come from general congressional legislation, and though the argument is novel, it is nonetheless worth making. As a matter of doctrine and policy, civil rights testing can fit under certain aspects of the First Amendment, and recognizing that would strengthen testers’ ability to engage in the truth-revealing work that they do while also protecting information-gathering tactics that crucially facilitate protected speech acts.
The First Amendment, with its protections for speech, expression, and the press, has been widely considered a hallmark of democracy. In order for it to serve its purposes, the First Amendment must be interpreted broadly.
And to that end, testing activity, as a method of investigating discriminatory practices, should be protected by the Constitution. The role of testers, especially civil rights testers, in uncovering legal violations is of the utmost importance to achieving the aims of a right to free speech and press: “‘A critical, independent, and investigative press is the lifeblood of any democracy.’”
Whether intentional or not, algorithms can function in discriminatory ways. Such discrimination can violate civil rights laws and carry great consequences for society. The CFAA, based on courts’ current interpretation, penalizes activities prohibited by websites’ terms of service and effectively blocks researchers, journalists, and civil rights advocates from conducting online civil rights testing. In doing so, the CFAA comes into tension with the First Amendment, which can be read to protect civil rights testing. To remedy this problem, rather than amending the CFAA or creating a separate statutory protection for civil rights testers, courts should acknowledge that the First Amendment does protect testing conduct and invalidate the current interpretation of the CFAA that prohibits online civil rights testing.