The field of machine learning and artificial intelligence is making rapid progress. Many people are starting to ask what a world with intelligent computers will look like. But what is the ratio of hype to real progress? What kinds of problems have been well solved by current machine learning techniques, which ones are close to being solved, and which ones remain exceptionally hard?
There isn’t currently a good single place to find the state of the art on well-specified machine learning metrics, let alone the many problems in artificial intelligence that are still so hard that there are no good datasets and benchmarks to keep track of them yet. So we are trying to make one. Today, we’re launching the EFF AI Progress Measurement experiment, and encouraging machine learning researchers to give us feedback and contribute to the effort.
We have drawn data from a number of sources: blog posts that report on snapshots of progress; websites that try to collate data on specific subfields of machine learning; and review articles. Where those sources didn’t have coverage, we’ve gone to the research literature itself and gathered data.
What we have thus far is an experiment, and we’d like to know: Is this information useful to the machine learning community? What important problems, datasets, and results are we missing?
EFF’s interest in AI progress is primarily from a policy perspective. We want to know what types of AI we need to start engaging with on legal, political, and technical safety fronts. Beyond that, we’re also just excited to see how many things computers are learning to do over time.
Given that machine learning tools and AI techniques are increasingly part of our everyday lives, it is critical that journalists, policy makers, and technology users understand the state of the field. When improperly designed or deployed, machine learning methods can violate privacy, threaten safety, and perpetuate inequality and injustice. Stakeholders must be able to anticipate such risks and policy questions before they arise, rather than playing catch-up with the technology. To this end, it’s part of the responsibility of researchers, engineers, and developers in the field to help make information about their life-changing research widely available and understandable. We hope you’ll join us.
EFF has just launched the Summer Security Camp, a two-week membership drive that challenges people everywhere to gather ‘round the online rights movement and prepare for the privacy and free speech challenges in their paths.
Through the 4th of July, anyone can join EFF or renew as a Silicon level member for just $20 and receive a set of miniature field guides with shareable security tips covering these cruciallyrelevantissues:
Border Search: know your rights and defend personal data at the border.
The EFF site contains extensive analysis of these topics and much more, but the Summer Security Camp's printed pocket guides distill some of the most important information to help keep you safe on the go, come what may. Members will have access to home-printable versions of these tips to share with friends and family because as we know, privacy is a team sport and everyone wins.
As a bonus, participants will receive a special edition embroidered patch to help them show support for the cause. Think of it as a digital civil liberties merit badge.
Threats to privacy and free expression abound, but EFF doesn’t believe in the no-win scenario. We work every day to defend user rights and empower you with knowledge that you can share in your community. The more prepared we are and the more we can count on each other, the stronger we’ll be. Let’s take a stand for online rights today!
The Supreme Court’s unanimous decision in Matal v. Tam striking down the trademark non-disparagement requirement as unconstitutional is a big victory for the First Amendment. First, the Court strongly pushed back against the expansion of the government-speech doctrine, perhaps the biggest current threat to free speech jurisprudence. Second, the Court strengthened a position EFF has long advocated—that intellectual property rights and First Amendment rights must be balanced against each other rather than weighted in favor of the former.
The case arose when the band The Slants was denied a federal trademark based on afederal law that prohibits the registration of a trademark that may “disparage. . . or bring into contemp[t] or disrepute” any “persons, living or dead.” The Court found that provision violated the First Amendment. It may no longer be used as a basis for denying trademark registration.
Pushing Back on the Dangerous Government-Speech Doctrine
The Governments’ primary argument in defense of the disparaging trademark ban was that registered trademarks were “government-speech,” not the speech of the trademark owner. That is, in denying registration, the government was not punishing The Slants because it disagreed with the viewpoint the mark expressed; rather, the government was simply choosing not to include disparaging terms in its own speech.
The government-speech doctrine is unique among First Amendment law in that it is the only situation in which the government may discriminate on the basis of the speaker’s viewpoint. In its most basic application, it is noncontroversial: the government itself may adopt policy positions and promote them without having to equally promote opposing policies advocating the opposite viewpoint. In all other contexts, the government cannot deny a speaker access to a forum or otherwise punish them because of a disagreement with the views expressed.
As the Court recognized in Matal, the government-speech doctrine “is susceptible to dangerous misuse. If private speech could be passed off as government speech by simply affixing a government seal of approval, government could silence or muffle the expression of disfavored viewpoints. For this reason, we must exercise great caution before extending our government-speech precedents.”
Significantly, the Court put a stop to what many saw as a gradual expansion of the government-speech doctrine through its previous decisions. The Court characterized its most recent government-speech decision,Walker v. Texas Div., Sons of Confederate Veterans, Inc., in which it held that a state’s specialty license plate program was government-speech, as “likely mark[ing] the outer bounds of the government-speech doctrine.”
The government-speech doctrine is unique among First Amendment law in that it is the only situation in which the government may discriminate on the basis of the speaker’s viewpoint.
The Court thus resoundingly rejected the government’s argument in Matal, explaining that it “would constitute a huge and dangerous extension of the government-speech doctrine.” It characterized the government’s position as “far-fetched” and not even “remotely support[ed]” by any of the Court’s previous government-speech decisions. Trademark registration does not bear any of the hallmarks of government-speech. Rather than articulating an official position by registering various trademarks, often of conflicting views, “the Government is babbling prodigiously and incoherently.” Moreover, “[t]rademarks have not traditionally been used to convey a Government message” and “there is no evidence the public associates the contents of trademarks with the Federal Government.”
Also highly significant to First Amendment doctrine, a plurality of the Court limited another aspect of its government-speech jurisprudence. In several cases, the Court has held that speech by private speakers but subsidized by the government may also be government speech and thus the provision of the subsidy may be subject to viewpoint discrimination without offending the First Amendment. But in Matal, the four justices rejected this argument and sharply limited these subsidy cases to those in which in the government makes cash payments for speech, not any other kind of subsidy.
Reasserting a Better Balance Between Free Speech and Trademark Law
The Court also reaffirmed that trademarks are expressive and imbued with First Amendment protections.
Perhaps the most worrisome implication of the Government’s argument concerned the system of copyright registration. If federal registration makes a trademark government speech and thus eliminates all First Amendment protection, would the registration of the copyright for a book produce a similar transformation? The justices unanimously rejected the government’s suggestion that trademarks could be distinguished from copyright on the ground that they are not expressive:
The Government attempts to distinguish copyright on the ground that it is “‘the engine of free expression,’” Brief for Petitioner 47 (quoting Eldred v. Ashcroft, 537 U. S. 186, 219 (2003)), but as this case illustrates, trademarks often have an expressive content. Companies spend huge amounts to create and publicize trademarks that convey a message. It is true that the necessary brevity of trade- marks limits what they can say. But powerful messages can sometimes be conveyed in just a few words.
In addition, the Court explained that the government does not have a greater ability to discriminate against disfavored viewpoints in registering trademarks merely because trademarks are “commercial speech.” Although commercial speech in many contexts gets somewhat diminished First Amendment protections, even commercial speech is not subject to the government’s viewpoint discrimination.
The U.S. Supreme Court, in Packingham v. South Carolina, unanimously struck down a state law that banned registered sex offenders (RSOs) from using all Internet social media, holding that the law violated the First Amendment.
EFF and our allies Public Knowledge and the Center for Democracy & Technology filed an amicus brief urging this result. The Court cited our brief for three propositions regarding the extraordinary consequences of banishing people from all Internet social media:
Seven in ten American adults use at least one Internet social networking service.
One of them, Facebook, has 1.79 billion active users.
All Governors and nearly all members of Congress use social media to communicate with their constituents.
The Court also cited our brief for the proposition that the broadly worded law might bar access not just to commonplace social media websites, but also to other websites like Amazon.com, Washingtonpost.com, and Webmd.com. Our brief was written by Professor David Post, as well as Jonathan Sherman, Perry M. Grossman, and Henry Bluestone Smith of Boies, Schiller & Flexner LLP.
Both Justice Kennedy’s majority opinion and Justice Alito’s concurrence in the judgment assumed without deciding that the law was content neutral, and thus applied the intermediate scrutiny test used for content neutral laws. Both opinions therefore required the government to prove that the law was narrowly tailored, meaning the law does not burden substantially more speech than necessary to achieve the government’s goal of protecting children. Both concluded that the law failed this test, because it banished RSOs from all Internet social media.
Several statements from the Court’s opinion (which Justice Alito’s opinion did not join) will be critical in deciding all manner of future cases applying the First Amendment to the Internet:
“Cyberspace . . . in general” and “social media in particular” are “the most important places (in a spatial sense) for the exchange of views.”
Internet social media “can provide perhaps the most powerful mechanism available to a private citizen to make his or her voice heard.”
“Even convicted criminals—and in some instances especially convicted criminals—might receive legitimate benefits form these means for access to the world of ideas, in particular if they seek to reform and to pursue lawful and rewarding lives.”
In addition to opposing the banishment of RSOs from all Internet social media, EFF also has long opposed government efforts to strip RSOs of their right to anonymousspeech on the Internet, and efforts to force RSOs to wear location-tracking shackles every moment for the rest of their lives.
EFF opposes laws like these that burden the digital liberties of RSOs for three reasons. First, digital liberty is a fundamental human right that all people should enjoy. Second, government often imposes new technological burdens on “the worst of the worst,” and then expands those burdens to other populations. Third, the government has designated nearly one million people as RSOs, including many non-dangerous people.
The Court’s decision in Packingham strengthens the First Amendment rights of all people to participate in the Internet.
Californians now have a chance to reclaim crucial online privacy protections.
Earlier this year, Congress narrowly voted to repeal federal privacy rules that kept your ISP from selling information about who you are and what you do online without your permission. Today, California legislators are introducing new state legislation—the California Broadband Internet Privacy Act, A.B. 375 (Chau)— that would effectively reinstate those rules for Internet users in California.
ISPs are our gatekeepers to the Internet, and we shouldn’t have to sacrifice our privacy to these companies just to get online.
The wildly unpopular vote in Congress earlier this year undid years of work at the FCC to create online privacy rules that codified and expanded on long-standing privacy protections. The updated rules, set to go into effect in late 2017, were necessary to protect personal information revealed to your ISP.
Without these privacy protections, ISPs like Comcast, AT&T, and Verizon—companies that you already pay to access the Internet—will have free rein to make even more money off of you by selling information about what you look at, what you buy, who you talk to, and more, online. And that ability to sell your information without your permission will open the door to measures that further harm your privacy and security on the Internet.
Because of the tool Congress used to repeal the FCC’s online privacy rules—called a Congressional Review Act resolution—the FCC can’t write similar rules in the future. And thanks to the current legal landscape, there’s no other federal agency that can protect Internet users from privacy violations by their ISPs.
That means state legislatures are the best place for Internet users to fight to reinstate their privacy rights. Eighteen other states have already taken steps to consider similar measures with Oregon’s HB 2813 (Williamson, Clem, Sanchez, and Marsh) being the most recent bill to be introduced.
The ISPs fought hard—using debunked industry talking points—to win in Congress. And since the rules were repealed at a federal level, ISPs have “pledged" to protect their customers’ privacy. But we know that their promises leave plenty of room for them to sell your information.
With A.B. 375, we have a chance to protect our privacy from ISPs’ privacy violations in California.
UPDATE (6/20/17): An earlier version of the post incorrectly identified the number of states with broadband privacy measures. In addition to California, eighteens states have introduced or considered broadband privacy legislation.
The U.S. government’s foreign surveillance law is so secretive that not even a service provider challenging an order issued by a secret court got to access it.
That Kafkaesque episode—denying a party access to the law being used against it—was made public this week in a FISC opinion EFF obtained as part of a FOIA lawsuit we filed in 2016.
The opinion [.pdf] shows that in 2014, the Foreign Intelligence Surveillance Court (FISC) rejected a service provider’s request to obtain other FISC opinions that government attorneys had cited and relied on in court filings seeking to compel the provider’s cooperation.
The decision was related to the provider’s ultimately unsuccessful challenge to a surveillance directive it received under Section 702, the warrantless surveillance authority that is set to expire this year.
The decision is startling because it demonstrates how secrecy jeopardizes one of the most fundamental principles of our justice system: everyone gets to know what the law is. Apparently, that principle doesn’t extend to the FISC.
The provider’s request came up amid legal briefing by both it and the DOJ concerning its challenge to a 702 order. After the DOJ cited two earlier FISC opinions that were not public at the time—one from 2014 and another from 2008—the provider asked the court for access to those rulings.
The provider argued that without being able to review the previous FISC rulings, it could not fully understand the court’s earlier decisions, much less effectively respond to DOJ’s argument. The provider also argued that because attorneys with Top Secret security clearances represented it, they could review the rulings without posing a risk to national security.
The court disagreed in several respects. It found that the court’s rules and Section 702 prohibited the documents’ release. It also rejected the provider’s claim that the Constitution’s Due Process Clause entitled it to the documents.
The opinion goes on: “Beyond what is compelled by the Due Process Clause, the Court is satisfied that withholding the Requested Opinions does not violate common-sense fairness.” This was because the Court believed that the DOJ had accurately represented the rulings in its legal briefs and did not mislead the provider about what those rulings said.
The court also said that even if the opinions were released, they “would be of little, if any assistance” to the merits of the provider’s arguments.
The court’s opinion notwithstanding, there is nothing fair about withholding important legal cases—which likely interpreted or created law—from one side in a legal dispute.
The court’s decision is akin to allowing one party to read and cite to a Supreme Court case while prohibiting the other side from doing the same. It fundamentally disadvantages one side in a legal fight, on top of denying it access to the case to ensure that the party in the know is accurately representing the ruling.
In the case of the provider, the deck was always stacked against its ability to challenge the 702 order. The FISC traditionally only hears from one party—the Executive Branch—and is usually sympathetic to claims of national security.
Although recent changes to the FISC as a result of USA Freedom Act have moved in the right direction, including the ability for outside parties to argue before the court, the DOJ still has many advantages.
In the case of the provider, the trump card was that the DOJ’s lawyers got to read and rely on cases that the provider never got to see.
To be sure, the unjust result is not entirely the fault of the FISC. As the ruling points out, Congress has provided little to no recourse for a party challenging secret surveillance orders to be able to obtain documents and FISC rulings that are directly relevant to its case.
With Section 702 due to sunset this year, Congress should recognize that the court system it set up to approve surveillance orders and hear challenges to those orders bears little resemblance to our broader justice system. This inequity corrupts our fundamental democratic principles and is yet another reason Congress must end Section 702.
On June 1, people from across the San Francisco Bay Area gathered for EFF's inaugural Tech Trivia Night under the watchful eye of quizmaster and Staff Technologist Cooper Quintin a.k.a. The Cybertiger. Following in the footsteps of EFF's popular Cyberlaw Trivia Night, now in its tenth year, this pub-style quiz featured a challenging array of questions geared toward developers, security engineers, and fans of obscure geek knowledge.
EFF's Cooper Quintin hosting the Famous Passwords round. 🐯
EFF organized the evening's gantlet of questions in six rounds: Bastard Operator from Hell, Rodentia, Famous Passwords, Name that Vuln, Cyber Cyber Cyber, and General Geekery. Trivia called upon a wide range of knowledge that stumped even some of the biggest of brainiacs. For example, the Rodentia Round (named for its rodent theme) included the question: What hypertext protocol created at the University of Minnesota has been assigned TCP port 70 by IANA?* This year's event even included questions about some famous cats of the Internet for good measure.
Senior Staff Technologist Jeremy Gillula, Staff Attorney Andrew Crocker, and Staff Technologist Erica Portnoy preside with the Gavel of Justice.
The twelve teams competed valiantly, but in the end it was a band of digital rights do-gooders known as The Labcoats who reigned supreme, receiving the First Place Tech Trivia Cup and an EFF swag superpack. Second Place honors went to Randos in the Light followed closely by Riven Mirror. Congratulations to the esteemed winners!
First Place Winners: The Labcoats
EFF would like to express our gratitude to Tech Trivia Night’s sponsors for helping make the first Tech Trivia Night a success. Many thanks to Facebook, Gandi, and Sonic! If you or your company are interested in supporting a future EFF event, please contact Nicole Puller. We can't wait to do it again.
EFF's sincere appreciation goes out to of the participants who joined us for a great quiz over dinner and drinks while never losing sight of our mission to drive the online rights movement forward. We salute the digital freedom supporters around the world who have helped ensure that EFF can continue working in the courts and with policymakers, activists, technologists, and the public to protect online privacy and free expression.
June 15, 2017 is the 100th anniversary of the passage of the Espionage Act. Earlier this year, as if to commemorate the centennial, President Trump suggested to then-FBI Director James Comey that he extend the Act into new territory—that he use it to prosecute journalists.
Unfortunately, there is little in the language of the law itself that would limit the application of the Espionage Act’s nondisclosure and possession prohibitions to journalists who did not themselves remove the documents from the control of the government. By their terms, two key sections of the Act seem to threaten criminal liability for journalists who shed light on government secrets. 18 U.S.C. § 793(e), for instance, broadly criminalizes the following:
Whoever having unauthorized possession of, access to, or control over . . . information relating to the national defense, which information the possessor has reason to believe could be used to the injury of the United States or to the advantage of any foreign nation, willfully communicates, delivers, [or] transmits . . . the same to any person not entitled to receive it, or willfully retains the same and fails to deliver it to the officer or employee of the United States entitled to receive it;
A related provision, 18 U.S.C. § 798, reaches anyone “who knowingly and willfully communicates . . . or otherwise makes available to an unauthorized person, or publishes, or uses in any manner prejudicial to the safety or interest of the United States” certain classified information.
These provisions have no express exception for journalists reporting on issues of public concern—and no implied shield has emerged from the handful of precedents handed down over the last century. The Supreme Court brushed up against the question for the first and only time in New York Times Co. v. United States, the federal government’s effort to block the publication of the Pentagon Papers. While the court’s brief per curiam opinion didn’t touch the issue, several justices wrote separately to suggest that the Justice Department could still prosecute the reporters responsible for the story after it made the front page.
Why have no such prosecutions followed? In part, probably, because the law is unclear, shot through with undefined terms that courts rarely see an opportunity to interpret. Justice Harlan called the Espionage Act “a singularly opaque statute” in his dissent from New York Times Co.; Professors Harold Edgar and Benno Schmidt, in a 1986 law review article, went so far as to call the statute “incomprehensible if read according to the conventions of legal analysis of text, while paying fair attention to legislative history.” There’s no guarantee that the prosecution of a journalist under the Act would succeed—and from the government’s perspective, a loss would undermine the coercive value of threatening prosecution to protect secret surveillance programs.
There are at least two compelling (but untested) legal arguments that would complicate any effort to prosecute a reporter who relied on leaked documents in that kind of case. In New York Times Co., the lower court thought—and Justice Douglas agreed—that the word “communicates” in § 793(e)’s was never intended to include publication by the media. That question is still an open one. What’s more, at least some Espionage Act precedents suggest that “reason to believe [the leaked information] could be used to the injury of the United States” is a stiff requirement, one that it’s hard to imagine would be satisfied in the case of public reporting. In United States v. Rosen, for instance, the district court for the Eastern District of Virginia insisted that that language “requires the government to demonstrate the likelihood of the defendant’s bad faith purpose to either harm the United States or to aid a foreign government.” In other words, it’s not enough to publish sensitive information that a news organization knows might harm national security; harming U.S. national security needs to be the point of publishing.
In essence, for decades, journalists have been protected against prosecution under the Espionage Act by an informal truce, not a battery of legal arguments.
The application of the disclosure provisions of the Espionage Act to third party recipients and republishers of government materials is likely unconstitutional under the First Amendment. Bartnicki v. Vopper, a 2001 case in which the Supreme Court afforded near absolute protection to the republication of information pertaining to matters of public interest that the republisher legally acquired even if their source illegally acquired and transmitted it to them. But the interaction of the Bartnicki doctrine and the Espionage Act has never been tested.
In essence, for decades, journalists have been protected against prosecution under the Espionage Act by an informal truce, not a battery of legal arguments. As Justice White noted—disapprovingly—in New York Times Co., a legal carve-out for reporters would “conform with the past practice of using the statute only to prosecute those charged with ordinary espionage.” Or as David Pozen described the government’s approach to leak prosecutions in The Leaky Leviathan,“In formal terms this legal regime looks forbidding, draconian. In practical terms, as a frustrated intelligence professional once put it, the system amounts to ‘permissive neglect.’”
That norm may be on the brink of changing. In 2014, Attorney General Eric Holder promised, “As long as I am attorney general, no reporter who is doing his job is going to go to jail.” Attorney General Jeff Sessions declined to make the same basic commitment at his confirmation hearings earlier this year. For better or worse, the Trump Administration may finally test the Espionage Act’s sweeping, ambiguous language against reporters’ core First Amendment rights.
With that test looming and with the 100th birthday of the law upon us, it’s clear that the Espionage Act needs to be reformed. Sign our petition calling for changes to this overly broad and outdated law.