Welcome back for another week of What's Next, where we report on the intersection of law and technology. This week, Clare Garvie of the Georgetown Law Center On Privacy and Technology breaks down the FBI and ICE's usage of facial recognition in DMV databases. Plus, lawyers from Littler Mendelson prepare employment lawyers for life in a world with deepfakes. And Congress calls the tech world in to discuss antitrust and censorship. Let's chat: Email me at [email protected] and follow me on Twitter at @a_lancaster3.

Face-to-Face with Law Enforcement's Use of AI

Last week, a Washington Post report revealing how the Federal Bureau of Investigation and U.S. Immigration and Customs Enforcement perform facial recognition searches of department of motor vehicle databases, ignited a debate on how the technology should be used in law enforcement, if at all.

The article relied on public records requests from Georgetown Law's Center on Privacy and Technology to show that the FBI alone does 4,000 facial recognition searches a month, many of which go through state DMVs. Clare Garvie, senior associate for the Georgetown program has read about 20,000 pages of records herself. After seeing that ICE conducts these searches, she set out to see if states had made formal agreements or laws to codify the use of facial recognition by the federal agencies.

In Washington, Utah and Vermont, the three states Georgetown Law Center focused on, there were no formal arrangements. “In looking at the laws in those states, what I found was that these are all states that not only purport to protect the rights of undocumented immigrants, but they also affirmatively allow undocumented immigrants to receive a driver's licenses or driver's privilege card in the name of public safety and extending a privilege to these people,” Garvie said. “So to turn around and allow ICE access to that same information, to me, is an outrage. It is a betrayal of trust that is very fundamental and feels very wrong.” 

About half of U.S. states have laws restricting their agencies from sharing DMV information with the federal government after the passage of the Real ID Act in 2005, Garvie said. “States like Utah, Arizona and others passed laws essentially saying they would not share this information with the federal government, because it overstepped the constitutional divide between states and the federal government,” she said.

Garvie said states such as Utah have began investigating how facial recognition is used in their DMVs. Over the last few years, Washington and Vermont have shut down DMV information sharing after learning about ICE and the FBI's general access, she said.

After seeing thousands of arrests made based on facial recognition, Garvie advocates for a moratorium on the use of the technology by law enforcement until legislatures approve the action. “In almost none of these cases was the defendant made aware that facial recognition was used,” she said. “Facial recognition is not a very accurate tool. It produces information that directly speaks to a defendant's guilt or innocence. Under our right to due process, this must be turned over to the defense, and it's not.”

Aaron Crews and Natalie Pierce of Littler Mendelson.

Deepfake in the Workplace

A few weeks back, the world watched a deepfake video of Facebook founder Mark Zuckerberg talking about how he holds our online lives in his hands. The video reportedly did not fool many people for long. Just imagine, said Aaron Crews, chief data analytics officer for Littler Mendelson, if the video was of Zuckerberg saying Facebook will no longer hire African Americans. Crews and Natalie Pierce, co-chair for Littler's robotics, AI & automation practice group, co-authored a report sharing how labor lawyers can hunker down for the oncoming tidal wave of deepfakes. Here's what you need to know about the technology that could spark a sea change in labor litigation.

➤➤ Can you describe a workplace scenario involving Deepfakes that could become more common in the future?  Aaron: There could be an employer-employee issue where there's a sex-related scandal arrising out of a deepfake video showing inappropriate conduct, and people get terminated immediately. That is a recipe for disaster in a world like this. And there are a lot of situations where these could be the catalyst for major class action litigation. Let's say there's a class action suit that says the company fires all of their workers over 40. And let's say there's a deepfake video on YouTube that is a representation of a major executive in the company that says, “Let's get rid of all of those older people, because they're not innovative enough, and they don't move as fast as our business.” There now has to be a real question of is this too good to be true?

➤➤ How should lawyers advise their clients to prepare for a world in which deepfakes are more common? Natalie: Like any emerging technology, it's so critical for leaders in an organization to increase awareness. In the same way you have a crisis communication plan, you have to have some sort of plan in place to respond when there's harmful and questionable content. We really are at a point where we need to teach people that just because you hear your CEO saying, “The robots are coming in from China next week, and we are going to be announcing the next mass layoffs,” doesn't mean it's true. It's also important to let employees know that you're taking measures, such as using cryptographic key signing, for example.

➤➤ How does cryptographic key signing work?  Natalie: It's just an additional method of verification. You can think of it as fingerprinting content at the video frame level from the time of capture through dissemination. On a simplistic level, it's similar to how DocuSign verifies signatures. It's an imperceptible watermark on a file.

Aaron: Essentially, you are holding up a private key that unlocks one side, and you create a complex equation that marks the document and carries with it at the file level. So if you record a press event where executives are speaking, before that's put out for public consumption, it could be marked cryptographically. If anyone wondered if it were real, they could use the key to verify.

➤➤ What are some other ways to detect deepfakes? Natalie: Right now, a lot of artificial intelligence scientists are developing technologies to detect videos with dropped frames. In connection with the upcoming 2020 election, they're looking at the pattern of how candidates are speaking and how that matches with their known body language.

Aaron: It's also an arms race. There are companies out there who are building technology in the same way companies started building antivirus software years ago. Those technologies identifying deepfakes are going to get better, but as we do that, deepfake technology is going to morph and improve. When I say this is an arms race, I really mean it in a nuclear proliferation sense. The detection technology is likely going to make better deepfakes if people reverse engineer it and deploy it in the algorithm used to make deepfakes.

➤➤ How could deepfakes change how we practice employment law?  Aaron: Realistically, in some ways, it's similar to how we realized eyewitness testimony is not necessarily the gold standard we thought it was. Historically, we've thought of video and audio as a gold standard. Now, we're several steps into a world where seeing is not necessarily believing. Recorded perception is now something that is up for debate among experts. That's going to be true in videos of electoral candidates, people who run companies and individual employees, and what they might have alleged to have said or done. There's some infrastructure and muscle that we're going to need to build in order to have that discussion in real time.

Google's Adam Cohen, Facebook's Matt Perault, Amazon's Nate Sutton and Apple's Kyle Andeer. Diego M. Radzinschi/ALM

Big Tech Gets Called to Washington

Silicon Valley tech companies headed to Washington this week to answer Congress' burning questions on antitrust activities and censorship.

On Tuesday, Facebook and Amazon bore the brunt of a House Judiciary Committee interrogation. Amazon, Facebook, Apple and Google testified for the committee on the cornering of their respective markets.

Rep. David Cicilline, D-Rhode Island, asked Nate Sutton, Amazon's associate general counsel for competition, about whether the online retailer uses data from its third-party sellers to promote Amazon's brands, citing that analysts estimate between 80% and 90% of sales go to Amazon's suggested “buy box” items.

“Our algorithms such as the buy boxes aim to predict what customers want to buy, and we apply the same criteria whether you're a third-party seller or Amazon, because we want customers to make the right purchase regardless of whether its a seller or Amazon,” Sutton said.

In light of antitrust actions swirling around Facebook, Rep. Jamie Raskin, D-Maryland, asked Matt Perault, Facebook's head of global policy development, if the company's plan to integrate Facebook, and its What's App and Instagram subsidiaries was a ploy to preempt efforts to break up the company.

“No, it is not,” Perault said. “We face fierce competition with the products and services we offer. It is our understanding that there are products on the market that offer more privacy protected services, so our pivot toward privacy in respect to interoperating our services was because of the competition we face on the market.”

Rep. Joe Neguse, D-Colorado, fired off active user stats at Perault, asking him to confirm that Facebook owns the majority of the top social media sites: “You can understand skepticism, because when a company owns four of the largest six entities measured by active users in the world in that industry, we have a word for that: monopoly, or at least monopoly power.”

Meanwhile in the other chamber, the Senate Judiciary subcommittee, lead by Sen. Ted Cruz, R-Texas, grilled Google representatives on their censorship practices around conservative speech.

“When you submit a video, people at YouTube determine whether you've engaged in hate speech, an ever-changing and vague standard meant to give censorship an air of legitimacy,” Cruz said. “This is a staggering amount of power to ban speech, to manipulate search results, to destroy rivals and to shape culture.”

However, Google denied the alleged anti-conservative tilt.

“Google needs to be useful for everyone, regardless of race, nationality or political leanings,” said Google's vice president for government affairs and public policy, Karan Bhatia. “We have a strong business incentive to prevent anyone from interfering with the integrity of our products, or the results we provide to our users. Our platform reflects the online world that exists.”

U.S. District Judge Michael Simon of the District of Oregon

ADR's Potentially “Really Messy Can of Worms”

In a lively hearing Friday, the U.S. Court of Appeals for the Ninth Circuit questioned how much information is enough for parties to decide the appearance of impartiality of arbitrators.

In Monster Energy vs. City Beverages, Energy drink company Monster Energy Co. was in court to confirm its arbitration award against distributor City Beverages, a.k.a., Olympic Eagle Distributing. However, Olympic says the award was unfair since its neutral is an owner of JAMS, the arbitration services provider used to resolve their dispute. As an owner, the neutral receives a share of profits from all JAMS cases, which Foster Pepper's Michael Vaska argued gives the arbitrator a stream of income from the 97 cases Monster has brought to JAMS in the last five years. Invoking the U.S. Supreme Court's decision in Commonwealth Coatings v. Cont'l Cas., Vaska said the neutral must disclose his role as an owner and his financial stake in Monster's repeat referrals.

U.S. District Judge Michael Simon of the District of Oregon, sitting in the case by designation, conceded that it might be really simple to disclose ownership status in an arbitration service provider. However, he also addressed Monster's concerns that it could expose arbitrators to a never-ending flood of subjective disclosure requirements.

“Can you give me any comfort at all that we're not opening a really messy can of worms that says based on Commonwealth Coatings if you have an ownership interest in an arbitration firm you must disclose that interest and the dollar value of repeat players?” Simon asked.

Vaska said California legislators have already looked into that can of worms and passed laws requiring disclosures about the arbitration service provider and his or her interest in consumer cases.

As more cases are tried outside the court system, Judge Milan Smith of the Ninth Circuit said that it's important to clarify disclosure rules.

“The Supreme Court was trying to give us some direction in connection with the FAA [in Commonwealth Coatings],” Smith said. “And as you know, the FAA kind of rules our lives these days. People need to understand how it works and what the rules are, because the arbitrators are doing more and more of the work that the courts used to do. The question is what rules apply to the arbitrator?”

On the Radar

Is Fee-Sharing Caring?  A state bar vote brought California one step closer to opening up its law firms to fee sharing with non-lawyers. The board of trustees for the organization voted in favor of accepting public comment on fee sharing and outside ownership of law firms. Although the stated goal of the initiative is importing fresh knowledge and funding into the industry, some lawyers say the move could invite the Big Four accounting firms into the market. Read more from Dan Packel here.

Crypto Contributions Although charities such as the Red Cross have started accepting cryptocurrency donations, evolving regulations and the fluctuating nature of digital assets could prevent their widespread acceptance in philanthropic giving. Legal experts are weighing in on the tax, infrastructure and logistic hurdles charities face in adopting cryptocurrency. Read more from Frank Ready here.

When Local Government Gets Hacked This summer, at least five municipalities have faced costly disruptions at the hands of hackers. Some jurisdictions coughed up $600,000 in bitcoin to regain control over their networks, while others had to kick it old-school with physical court filings for about six weeks. The recent intrusions prove that government, like private industry, has a lot to learn when it comes to cybersecurity. Read more from Victoria Hudgins here.

Thanks for reading. We will be back next week with more What's Next.