Showcasing our researchers’ work and their latest media appearances as they weigh in on developments at the intersection of national security and emerging technology.
Bloomberg broke the news that Dutch semiconductor manufacturer ASML Holding has accused a Beijing-based firm of intellectual property theft after successfully suing the firm’s predecessor. China, which lags in semiconductor manufacturing and is dependent on foreign technology, has taken steps to acquire advanced technology through a variety of means, including IP theft. “Taken individually these cases can seem really anecdotal, and some victim companies might say there’s not really any master plan,” said Senior Fellow Anna Puglisi. “But take this together with other cases over time and you can see the silent—and in some cases not so silent—hand of the Chinese government.”
The Wall Street Journal
Clamping down on facial recognition technology and data privacy violations will spur innovation, The Wall Street Journal reported. Facial recognition company Clearview AI was fined $10 million by the U.K.’s Information Commissioner’s Office for the collection of personal data without consent. However, this move is unlikely to hinder Clearview AI’s use of facial-recognition technology or its ability to expand, Research Analyst Dahlia Peterson told the Journal, because “fines that come after the fact may do little to stop image data exploitation.” According to Peterson, strict privacy protections in the U.K. and elsewhere in Europe have forced technology companies to build innovative software that balances with people’s right to privacy, such as developing automated face-pixelation capabilities for live video surveillance. Steps can also be taken to improve the accuracy of AI models that use synthetic biometric data rather than images of real people, Peterson noted.
The National Interest
In an opinion piece for The National Interest, Peterson argued that the United States and its allies should levy Magnitsky sanctions on Chinese AI surveillance giant Hikvision for its abuses against Uyghurs and other Turkic Muslims in Xinjiang internment camps. “In a move that would send a strong signal about how seriously committed the United States and its allies are to ending tech-enabled repression and genocide, these sanctions would be a well-founded warning to any company that intends to remain intertwined in the Xinjiang surveillance complex,” Peterson wrote. She also recommended that Hikvision be added to the Treasury Department’s Specially Designated Nationals (SDN) List through the Global Magnitsky Act to quickly implement sanctions and supercharge policy actions against the technology company. “Since other Chinese AI champions could act as substitutes for Hikvision in enabling repression in Xinjiang,” Peterson noted that “the [Biden] administration should warn against this by publicly signaling that it will be closely monitoring the impacts and evaluating the possibility of adding other major players to the SDN list in the future.”
Reducing U.S. sanctions against Chinese chip companies could reduce inflation and the likelihood of an economic war between the United States and China according to another opinion piece in The National Interest. It states that U.S. restrictions on Chinese chipmakers have led other chip manufacturers to overutilize their equipment and offshore production to keep up with demand. With the lagged semiconductor factory permitting process, CSET research found that “the United States currently builds fewer fabs at a slower rate than the rest of the world.”
In an opinion piece for The Hill, Research Analyst Will Hunt and former CSET Research Fellow Remco Zwetsloot explained that funding from the U.S. Innovation and Competition Act and the America COMPETES Act isn’t the only resource needed to bolster U.S. supply chains. “To give onshoring the best chance of success, the United States must also onshore the STEM talent it needs to compete,” which is in short supply, according to the authors. Drawing from their CSET research, Hunt and Zwetsloot estimated the staffing needs of the chipmaking facilities that would be built with funding from the new innovation legislation and found that U.S. chipmakers would need to hire tens of thousands of additional workers — many of them from overseas, where most of the world’s scientists and engineers with the necessary skills reside. Foreign-born STEM talent already makes up approximately 40 percent of all high-skilled workers in the U.S. semiconductor industry. Prior to today’s current talent shortages, a recent CSET report found that access to talent has been a key motivation for offshoring among U.S. semiconductor companies. According to the authors, STEM immigration reform would help retain the world’s best talent and onshore U.S. semiconductor manufacturing.
The continued innovation of large language models (LLMs) spurs development of new cutting-edge AI technology, but with innovation comes risk. To enforce better practices for the deployment of AI language models Cohere, OpenAI, and AI21 Labs composed a joint set of best practices and key principles for LLM providers to mitigate risk. In an article by VentureBeat, CSET alongside other organizations within the technology community was cited for their endorsement of the joint recommendations. Director of CyberAI John Bansemer expressed his support that “as large language models have become increasingly powerful and expressive, risk mitigation becomes increasingly important. We welcome these and other efforts to proactively seek to mitigate harms and highlight to users areas requiring extra diligence. The principles outlined here are an important contribution to the global conversation.”
Export Compliance Daily
In CSET’s May 23rd webinar, A New Export Control Regime for the 21st Century, Research Fellow Emily Weinstein and Non-Resident Senior Fellow Kevin Wolf presented their vision for a new export control regime among techno-democracies to better address contemporary challenges. Export Compliance Daily recapped their view that a new export control regime could keep sensitive technology from misuse by authoritarian governments and reduce pressure on the U.S. to impose unilateral control. A previous Export Compliance Daily article reiterated the researchers’ key findings from their May presentation. “The unilateral route is potentially effective in the short term, but not in the long term, and usually ends up hurting industry more,” Weinstein said. “We need to tackle these issues with allies not only just to stay friendly with allies, but to ensure that we are actually closing all of the gaps we need to close if we’re using export controls.” Wolf and Weinstein said they hope countries can “move as quickly as possible” to set up a new regime, particularly because the attention and resources devoted to Russia sanctions may fade over time. “Given the urgency of the issues and the difficulty of getting alignment, moving quickly is really our suggestion,” Wolf said. “The allies have shown they can move quickly when they want to.”
To understand how the Chinese government uses guidance funds to invest in emerging technologies, the journal American Affairs reached out to Research Analyst Ngor Luong. In her report “Understanding China’s Government Guidance Funds,” Luong alongside Zachary Arnold and Ben Murphy found that “guidance funds allow the Chinese state to leverage market discipline and resources. Chinese policymakers began to recognize the flaws of subsidy schemes and other traditional industrial policy tools. By bringing the profit motive into industrial policy, guidance funds aim to avoid these problems.” Luong noted that “often times, [guidance funds] don’t raise as much as intended, and they don’t invest the way they are supposed to because their SEO investors have no appetite for risk. Instead of hiring professional managers, they routinely recruit local officials. Guidance funds may actually crowd out other types of investments.” The Financial Times also took notice of Luong’s research on how guidance funds are “poorly conceived and implemented, and that the mechanism as a whole is often inefficient.”
Government CIO Media & Research
Defense leaders at the Department of Defense’s Digital & AI Symposium regard culture as a major hurdle in the department’s AI strategy, according to an article by Government CIO Media & Research. Deputy Director of Analysis Margarita Konaev expressed in a panel discussion that the DOD’s bureaucratic structure could impede a robust AI culture from being realized. “There are serious challenges that remain … [such as] a risk-averse culture, the preference to continue with the path we’re already in and the lack of incentives to promote AI and analytics and data and foster a more experimental culture,” said Konaev. The contrast is surprising in a sense because you have a military at the strategic and tactical level that is ready to learn and experiment from previous engagements and conflicts. At the same time at the organizational level, we have a lot more static structure that tends to favor the past as opposed to the future of war.”
Spotlight on CSET Experts: Dahlia Peterson
Dahlia Peterson is a Research Analyst focused on China and its AI surveillance capabilities. Prior to joining CSET, she researched how China harnesses predictive policing algorithms and facial, voice, and gait recognition technologies for its AI-powered surveillance programs.
Her recent publications include AI Education in the United States, Education in China and the United States, and China is Fast Outpacing U.S. STEM PhD Growth. Her work has been featured in a variety of outlets including Forbes, The Hill, Brookings, and Marketplace,
Interested in speaking with Dahlia or our other experts? Contact External Affairs Specialist Adrienne Thompson at firstname.lastname@example.org
Want to stay up to date on the latest CSET research? Sign up for our day-of-release reports and take a look at our biweekly newsletter, policy.ai.