UK “inclined to reject” Nvidia-Arm deal, Australia and South Africa recognize autonomous inventors, and the U.S. military tests AI and data sharing

Sign up
Worth Knowing

UK Leans Toward Blocking Nvidia-Arm Deal: The UK is considering blocking Nvidia’s $40 billion acquisition of British chip designer Arm, according to a recent Bloomberg report. As we covered earlier this year, the UK’s digital secretary directed the country’s competition watchdog, the Competition and Markets Authority, to prepare a report on the market and national security implications of the deal as part of a public interest intervention. That report, delivered to the digital secretary on July 20 but not yet made public, included “worrying implications for national security” that make the government “inclined to reject the takeover,” according to a Bloomberg source. The deal has attracted particular scrutiny because of concerns that it would jeopardize Arm’s status as “the Switzerland of chips” — Arm acts as a neutral licenser of chip designs to more than 500 companies worldwide, but many are among Nvidia’s biggest competitors. Now that the CMA’s report has been submitted, the digital minister can clear the deal, clear it with restrictions, or — as the Bloomberg report indicates is most likely — begin a more comprehensive “phase two” investigation.
Autonomous Inventors — International Legal Effort Yields Its First Wins: South Africa’s patent office awarded an AI system a patent for something it invented — a world first — while an Australian court ruled that AI systems can be recognized as inventors. Both cases were the result of the Artificial Inventor Project (AIP), an international campaign waged by a UK-based professor, Ryan Abbott, and a U.S. AI developer, Stephen Thaler, who created the AI “inventor” in question, named DABUS. The AIP filed patent applications in 17 countries listing DABUS as the sole inventor of a new type of food container based on fractal geometry. Until now, those applications had been unsuccessful — the U.S. Patent Office said that only “natural persons” could be recognized as inventors in its decision to reject DABUS’s application, and the UK IP Office ruled similarly (a decision that was later upheld in court). The Australian court’s ruling overturns a similar interpretation from IP Australia, but with observers anticipating an appeal, it could be some time before DABUS — or any other autonomous inventor — receives an Australian patent.  
China’s Top Court Limits Facial Recognition: The People’s Supreme Court, China’s top judicial body, ruled that the use of facial recognition by private businesses without consent is prohibited. With private facial recognition technology in relatively wide use — mobile phone companies, concert venues and property agents have been using the technology to monitor clients, for example — the ruling is expected to have far-reaching effects. But the decision, set to go into effect on August 1, is not an aberration: it comes in the midst of a broader Chinese government effort to rein in commercial tech companies and their use of private data, through such measures as the 2017 Cybersecurity Law, the upcoming Data Security Law (expected in September), and the Personal Information Protection Law (expected next year). These efforts have drawn comparisons to privacy efforts elsewhere, including the European Union’s General Data Protection Regulation, but while those movements have also taken aim at government use of surveillance technology, no such parallel effort exists in China.
Government Updates

U.S. Military Tests AI Capabilities and Data Sharing: Last month, DOD conducted the third in a series of Global Information Dominance Experiments (GIDE 3), a cross-command exercise meant to test the military’s data-sharing capabilities and the real-time use of AI systems. GIDE 3 — run by NORAD and Northern Command in collaboration with 11 U.S. combatant commands, the Joint Artificial Intelligence Center and others — tested JAIC’s “Matchmaker” tool, which enables “machine-enabled crisis deterrence and conflict defense options,” according to an Air Force press release. GIDE 1 and 2, conducted in December 2020 and March 2021, respectively, tested AI-enabled early warning systems and practiced cross-combatant command coordination. Observers say the series of experiments is a key step toward implementing Joint All Domain Command and Control, the broader strategy meant to accelerate AI integration across the military. While the military acknowledged preliminary plans for a GIDE 4, no date has been set.    

Senate Committee Advances Two AI Bills: Last week, the Senate Committee on Homeland Security and Governmental Affairs unanimously advanced two AI-related bills, the AI Training Act and the Deepfake Task Force Act. The former, sponsored by Sens. Portman and Peters, would establish a training program to educate the federal acquisition workforce — federal employees who might purchase AI systems for government use — about the benefits and risks of AI. Topics covered in the OMB-administered training would include how AI works, how to identify trustworthy AI systems and risks associated with AI. The Deepfake Task Force Act, meanwhile, would establish a task force in the Department of Homeland Security composed of government and industry experts and charged with developing a plan to counter deepfakes. While the bill has received support from some major tech figures, others argued that the national security threat posed by deepfakes has been overblown.

Members of Congress Push Back Against China Initiative: A bicameral group of more than 90 members of Congress sent a letter to Attorney General Merrick Garland questioning the Justice Department’s “China Initiative” and requesting an investigation into the “wrongful targeting of individuals of Asian descent for alleged espionage.” Launched under the Trump administration in 2018, the China Initiative aimed to crack down on China’s targeting of U.S. technology, including early stage research. While the initiative has continued under the Biden administration, it has faced significant pushback in recent months, especially after a handful of botched cases against individuals accused of spying for China, most notably that of Anming Hu. Last month, the DOJ dropped charges against five Chinese researchers accused of espionage, though officials insisted it wasn’t for lack of evidence and cited the deterrent effect of the cases — more than 1,000 researchers with ties to the PLA reportedly left the country after the arrests. The DOJ appears, for its part, unfazed — last week, prosecutors filed a notice in district court seeking a retrial in the Anming Hu case.  

In Translation
CSET’s translations of significant foreign language documents on AI

Shanghai Economic Reform Opinions: Opinions of the CCP Central Committee and the State Council on Supporting High-Quality Reform and Opening Up in Pudong New District and Making it into a Leading Area for Socialist Modernization Construction. These “Opinions,” made public by the Communist Party in July 2021, outline new policies for Pudong New District in Shanghai, long a trendsetter in Chinese economic reform. The document introduces several new measures to liberalize Shanghai’s capital market, including the STAR Market, where many Chinese AI companies are listed. The “Opinions” also strengthen Shanghai’s university- and research laboratory-based technology transfer agencies and call for aggressive recruitment of overseas tech talent.

If you have a foreign-language document related to security and emerging technologies that you’d like translated into English, CSET may be able to help! Click here for details.

What We’re Reading

Article: Mitigating dataset harms requires stewardship: Lessons from 1000 papers, Kenny Peng, Arunesh Mathur and Arvind Narayanan, arXiv (August 2021)

Article: View of Ethics and Governance of Artificial Intelligence: Evidence from a Survey of Machine Learning Researchers, Baobao Zhang, Markus Anderljung, Lauren Kahn, Noemi Dreksler, Michael C. Horowitz and Allan Dafoe, Journal of Artificial Intelligence (August 2021)

Article: Tortured phrases: A dubious writing style emerging in science. Evidence of critical issues affecting established journals, Guillaume Cabanac, Cyril Labbé and Alexander Magazinov, arXiv (July 2021)

What’s New at CSET

REPORTSPUBLICATIONS, PODCASTS AND EXPERT TESTIMONYFORECASTS
CSET maintains a crowd forecasting platform. Sign up as a forecaster, and take a look at some of the predictions so far:
IN THE NEWS
Schedule Update

policy.ai will be taking a brief summer break over the next few weeks, but don’t worry — we’ll be back on September 9 with all the latest in AI, emerging tech and security policy!