In his latest post for Avast, Garry Kasparov examines the intersection of privacy, transparency, security, human rights, and institutions in the age of AI.
I was honored recently to be one of the recipients of the Electronic Privacy Information Center’s (EPIC) 2017 Champions of Freedom award. Tech and privacy giant Bruce Schneier was among the presenters, and my fellow recipients—attorney Carrie Goldberg, Ron Rivest, and Judge Patricia Wald—are all doing important work in protecting privacy online and off in an age where it feels like we are always being watched. EPIC president and executive director Marc Rotenberg made a powerful statement about the vital, and often unsung, role that transparency has in a healthy democracy.
Users are concerned as well, and that is an understatement. The term “privacy” currently has 58,400,000 Google News mentions, enough to where it’s worth pointing out that the term means different things to different people in different frameworks. As EPIC well understands, the issue isn’t just end-user privacy and oversharing on social media; it’s also what happens at the top levels of the governments and corporations that have the capabilities to monitor our actions.
This is why privacy and transparency are two sides of the security coin. Companies need to know something about us to provide services we want. Government agencies need to monitor to provide essential security services. The eternal question is how to balance these needs with the rights of citizens to maintain their privacy and control over their own data even as we produce more of it every day. Transparency is required to establish limits—and accountability for overstepping them.
This adversarial process is part of the checks and balances that democratic republics are based on, especially in the United States. The idea is to use pressure and conflict to expose and fix the cracks in the system, to shine light and take gradual steps for the greater good. It’s important to contrast this with the nature of surveillance and privacy in the unfree world. The method, motive, and mission behind data collection and violations of privacy cannot be ignored. In authoritarian regimes, privacy is only for the rulers while the people have none. The behavior of the regime and its most powerful citizens is shielded from the people. Data collection is used for repression and persecution of innocent citizens. Real people are in real physical danger when privacy protections fail in an autocracy; it’s not a legal exercise, or done to protect people from hackers or terrorists. There could be no organization like EPIC in Putin’s Russia—and if there were, it would just be another branch of the security services.
Our desire to maintain privacy is inseparable from our concerns about the ethical underpinnings of our institutions. If we believe the motivations of authorities are morally sound, and operating in our interest, and the interests of society, we are more comfortable relinquishing a degree of privacy. You might call this “watching the watchers;” those who are in charge of establishing surveillance protocols must be monitored themselves. Their justifications must be subject to scrutiny and there must be a chain of responsibility and accountability. Ideally, we want a system that allows for the collection of information needed for economic, security, sociological, and other purposes, but deters arbitrary collection that does not serve a specific and well-articulated need. Governments and corporations must be transparent in expressing their goals and strategies, and consumers have the duty to question any that make them skeptical.
I am, of course, always interested in how AI comes into the picture. We worry about internal decision-making processes in bodies that are hidden from the public eye. How can we be sure that the motivations giant corporation, deciding how much user information to disclose to foreign governments, has the best interests of its users in mind? If we worry about this, and rightly so, we ought to think carefully about how we will delegate similar tasks to AI in the future. Increasingly, algorithms are making decisions with weighty consequences for individuals, companies, and societies. With the field of machine learning growing at a rapid pace, we will often find ourselves in the position of having to evaluate the results of processes that are unknowable to humans. A neural net that reached a particular endpoint did so through a series of steps that its human designers did not engineer, but through self-coding. These trends are leading us toward a future of unpredictable results and even less clear chains of responsibility for those results.
Unsurprisingly, algorithmic transparency and accountability is a growing field in privacy law; EPIC has launched a campaign in this area. If we cannot trace the processes of an AI and ascertain how it reached a certain decision, we must at the very least make transparent all elements of the process that were controlled by humans. As I argued above, we must demand clear and morally-backed justification from the human institutions responsible for broad surveillance. The same is true here: we must demand that AI is programmed according to the most rigorous transparency and ethics standards.
There must also be a framework to determine responsibility—human responsibility. Where does the liability belong when an internet of things connected toaster joins a botnet that brings down a chunk of the internet? Is the manufacturer responsible for making an unsafe product? As with dealing with fake news and other symptoms of our powerful technology run amok, we have to do a better job of assigning sources and responsibility to people instead of blaming technology.
Government and industry standards can make a big difference, however. After the EPIC award ceremony on June 5, I had a brief chat with encryption pioneer Whit Diffie. He was dismayed by my comment that people should know better by now not to click on unsafe email attachments. He pointed out that it’s ridiculous that users have the option to do something so potentially destructive at all, and he has a point. Modern skyscrapers don’t have windows that can open more than a crack—to avoid accidents and suicides. You can never prevent people from self-inflicted damage completely, but there is clearly a lot that could be done to improve our digital structures in ways that would help protect us from ourselves.
The crucial link is institutions, public and private, and whether there exists enough transparency and accountability in them to trust them with our privacy and other forms of security. Whether we are thinking of a government making decisions about domestic surveillance programs or a company determining how its products will collect information from users, if the values and norms governing human (or machine) action are strong enough, we can be optimistic about reaching favorable outcomes—even if it comes after years of trial and error. If institutions are lacking, we make ourselves vulnerable to threats from nefarious actors or neutral technologies that can cause harm, either by accident or by malicious intent.
In order to be strong, institutions must have continuity across time, and that includes across changes in leadership and party. In today’s hyper-partisan landscape, the latter can be hard to stomach, but I pride myself on being a vocal critic of the hypocrisy that results from political extremism on both sides of the political spectrum. In my acceptance speech at EPIC, I emphasized that if you are okay with surveillance and executive orders when you like the president, but you hate surveillance powers and executive orders when you don’t like the president, then you are part of the problem! Building a system that works for all of its members requires compromise from all sides. A framework complete with common-sense transparency laws and robust checks and balances on monitoring authorities will never materialize from finger-pointing and retreating into the safe, but dead-end, corners of total privacy or unbridled surveillance.
In the meantime, as we work to build the institutions that will shape our personal experiences and interactions in the digital world, we must be responsible as individual users acting within an imperfect system. The online world is far from a civilized one. It’s more like the old American Wild West, where the laws are poorly established and frequently abused. The ideal safeguards do not exist, and so each consumer must be careful of the information he shares and the measures he takes to protect it. Each of us must continue to push for the laws and protections needed for collective security, while reconciling ourselves to the current reality and remaining vigilant.
The lines for accountability are still weakly drawn, so companies have little incentive to put protections in place. And as popular as it is to demonize “big government” as a road to “Big Brother,” and to target ever-expanding data collection by some of the world’s largest companies, these giant institutions are actually the most accountable ones we have. Larger companies with greater name recognition and a reputation to defend are obliged to concern themselves with public opinion and user satisfaction.
In the same way, many governments simply do not care about the optics of violating their citizens’ privacy while those that are concerned about their standing on the world stage will do better. Protect yourself as best as you can, using the most reputable tools available—the Colt 45s and Winchester rifles of the digital Wild West armory—but do not stop pushing for the broader institutional change that will lead to a safer, fairer, and more transparent online world for everyone—hopefully without a shot being fired.
The rise of online tribes requires new strategies for countering the spread of misinformation. This is because your entire identity is wrapped up in a tribe and once you're a member, leaving can come with dire consequences.
Defining criteria for creating the ideal "tech hate stack" on platforms including Parler, Facebook and Telegram.
Over time, computer malware has moved from being largely harmless to having significant real-world consequences. This transition revolves around money, as new profit avenues arose for hackers when more aspects of our lives migrated to the online world.