Internet censorship by authoritarian governments prohibits free and open access to information for millions of people around the world. Attempts to evade such censorship have turned into a continually escalating race to keep up with ever-changing, increasingly sophisticated internet censorship. Censoring regimes have had the advantage in that race, because researchers must manually search for ways to circumvent censorship, a process that takes considerable time. New work led by University of Maryland computer scientists could shift the balance of the censorship race. The researchers developed a tool called Geneva (short for Genetic Evasion), which automatically learns how to circumvent censorship.

Membership inference attacks seek to infer the membership of individual training instances of a privately trained model. This paper presents a membership privacy analysis and evaluation system, called MPLens, with three unique contributions. First, through MPLens, we demonstrate how membership inference attack methods can be leveraged in adversarial machine learning. Second, through MPLens, we highlight how the vulnerability of pre-trained models under membership inference attack is not uniform across all classes, particularly when the training data itself is skewed. We show that risk from membership inference attacks is routinely increased when models use skewed training data. Finally, we investigate the effectiveness of differential privacy as a mitigation technique against membership inference attacks. We discuss the trade-offs of implementing such a mitigation strategy with respect to the model complexity, the learning task complexity, the dataset complexity and the privacy parameter settings. Our empirical results reveal that (1) minority groups within skewed datasets display increased risk for membership inference and (2) differential privacy presents many challenging trade-offs as a mitigation technique to membership inference risk.

Health care innovators are developing artificial intelligence algorithms called Explainable AI (XAI) that actually reveal the logic behind their diagnoses. Because their results can be verified, doctors and regulators will be more likely to adopt these algorithms than traditional "black box" AI. However, the transparency that makes these algorithms valuable to practitioners also makes the technology trickier to protect as intellectual property. With some legal creativity, there are multiple paths to patent protection for XAI-based diagnostics. The very nature of XAI algorithms prevents them from being kept secret, and the law governing patents for diagnostic algorithms is nearly undecipherable.

Jaimi has worked with Brendan's team enough to know how much talent acquisition professionals are craving to hire candidates quickly, fairly, and efficiently. At the same time, there has been explosive growth in the number of software tools that have been brought to market to help recruiters do just that. "On LinkedIn, you'll see a bunch of ads," Jaimi says, "for a bunch of different vendors that say, 'Hey, we can use machine learning to help you find the right candidate.'" She notes that often these companies also tout that their products eliminate human bias. "It sounds great," Jaimi tells Brendan, "and it's not to say that it couldn't be great, but there can be really serious unintended consequences that can cause a legal liability."

Epiq has deep experience applying advanced analytics to eDiscovery matters for the benefit of its clients, using AI in over 1,000 matters in the past year, spanning an array of litigated matters, regulatory reviews, and internal investigations. Recent client engagements include leveraging AI for a leading international bank, a national healthcare provider, and one of America's largest cities. The matter type and project size differed greatly in each case, but those clients sought the cost, time, and quality advantages that only Epiq can offer. About Epiq Epiq, a global leader in the legal services industry, takes on large-scale, increasingly complex tasks for corporate counsel, law firms, and business professionals with efficiency, clarity, and confidence. Clients rely on Epiq to streamline the administration of business operations, class action and mass tort, court reporting, eDiscovery, regulatory, compliance, restructuring, and bankruptcy matters.

Time is money and, unfortunately for companies, hiring new employees takes significant time – more than a month on average, research shows. Hiring decisions are also rife with human bias, leading some organizations to hand off at least part of their employee searches to outside tech companies who screen applicants with machine learning algorithms. If humans have such a hard time finding the best fit for their companies, the thinking goes, maybe a machine can do it better and more efficiently. But new research from a team of Computing and Information Science scholars raises questions about those algorithms and the tech companies who develop and use them: How unbiased is the automated screening process? How are the algorithms built?

These programs are trained to recognize patterns that indicate that something is fishy with a particular email. These programs can analyze thousands of emails in seconds and quarantine those that look iffy. As they grow and learn more about the patterns in suspicious emails, they become even more effective. The same is true of the anti-virus software that protects all our data.

Artificial intelligence (AI) has generated increasing interest in "future of work" discussions in recent years as the technology has achieved superhuman performance in a range of valuable tasks, ranging from manufacturing to radiology to legal contracts. With that said, though, it has been difficult to get a specific read on AI's implications on the labor market. In part because the technologies have not yet been widely adopted, previous analyses have had to rely either on case studies or subjective assessments by experts to determine which occupations might be susceptible to a takeover by AI algorithms. What's more, most research has concentrated on an undifferentiated array of "automation" technologies including robotics, software, and AI all at once. The result has been a lot of discussion--but not a lot of clarity--about AI, with prognostications that range from the utopian to the apocalyptic.

Artificial intelligence technology is advancing and bringing opportunities for society but also profound challenges for individual freedom. AI is a powerful enabler of surveillance technology, such as facial recognition, and many countries are grappling with appropriate rules for use, weighing the security benefits against privacy risks. Authoritarian regimes, however, lack strong institutional mechanisms to protect individual privacy--a free and independent press, civil society, an independent judiciary--and the result is the widespread use of AI for surveillance and repression. This dynamic is most acute in China, where the Chinese government is pioneering new uses of AI to monitor and control its population. China has already begun to export this technology along with laws and norms for illiberal uses to other nations. As AI-enabled surveillance technology spreads around the globe, how it is used poses profound challenges for the future of democracy, liberty, and individual freedom.

The role of the development of artificial intelligence in geopolitics usually means competition between the United States and China. While reports are inconclusive about which country will ultimately win (if this is the right term), there is a glaring shortcoming in the United States that China will largely avoid. In 2017 China accounted for 48% of global AI venture capital while the US only accounted for 38%. But only two years later the trend reversed, possibly because of headwinds from the trade war. Neither is it talent acquisition and a brain drain.