Cyber Week in Review: April 19, 2024
from Net Politics and Digital and Cyberspace Policy Program

Cyber Week in Review: April 19, 2024

Four Chinese firms blacklisted by the United States; Samsung awarded semiconductor grant; misleading posts spread on X after drone attack; UK may prohibit explicit deepfakes; CAICT releases AI benchmarks.
An anti-missile system operates after Iran launched drones and missiles towards Israel, as seen from Ashkelon, Israel on April 14, 2024.
An anti-missile system operates after Iran launched drones and missiles towards Israel, as seen from Ashkelon, Israel on April 14, 2024. Amir Cohen/Reuters

Four Chinese firms blacklisted by the United States

The Department of Commerce announced that the United States is adding four Chinese companies to an export blacklist for attempting to skirt export controls to provide AI chips to the Chinese military. The blacklisted companies accused are Linkzol (Beijing) Technology Co, Xi'an Like Innovative Information Technology Co, Beijing Anwise Technology Co, and SITONHOLY (Tianjin) Co. Additionally, the Department of Commerce banned eight other companies that allegedly aided drone production for Russian attacks in Ukraine and Iranian attacks in the Red Sea. The U.S. government has attempted to stem the flow of advanced microchips into China for years. In October 2023, the Commerce Department expanded and tightened its export controls on China’s acquisition of advanced microchips. The October 2023 decision prevents China from buying manufacturing equipment and chips used in certain major AI processes, including Nvidia’s H800 and A800 chips, which had previously been designed to comply with existing controls. The Chinese government denied that the companies were breaching Commerce Department rules, with Foreign Ministry spokesperson Mao Ning stating that blacklisting the companies was an attempt to “contain and suppress” Chinese companies.

Biden administration awards Samsung $6.4 billion to build two semiconductor fabs in Texas

The U.S. Department of Commerce announced that the Biden administration had signed a nonbinding agreement for a $6.4 billion dollar grant with Samsung; the funding comes from the 2023 CHIPS and Science Act, which allocated $50 billion to encourage American semiconductor manufacturing. The grant will be used to grow Samsung’s research and development presence in Texas and expand existing production facilities for leading-edge chips U.S. Secretary of Commerce Gina Raimundo stated that the CHIPS Samsung investment would be a “catalyst for continued private sector investments to help secure the long-term stability we need to put America at the beginning of our semiconductor supply chain.” Samsung is expected to invest more than $40 billion in Texas, creating 17,000 construction jobs and 4,500 construction jobs in the process. In addition to the grant, Samsung has tentative plans to claim the U.S. Department of the Treasury’s Investment Tax Credit, which is expected to cover up to 25 percent of qualified capital expenditures associated with the plants.

Misleading and deepfake posts about Iran’s drone and missile attack go viral on X

More on:

Artificial Intelligence (AI)

Digital Policy

Iran

Israel

The Institute for Strategic Dialogue (ISD) found that posts containing misleading and/or outright fake information about Iran’s drone and missile attack against Israel on April 13 were going viral on X. The report found that misleading content related to Iran’s attack had been viewed over thirty-seven million times. The misleading posts contained a variety of different content, including AI-generated images of drones and missiles, video game footage, and footage from other conflicts that was then misattributed to the current strikes. Many of the posts from the ISD report were shared widely by accounts that had paid for account verification, which increases content amplification through X's algorithms. Only two of the misleading posts had been tagged with a  community note, a type of content moderation feature that allows users to flag misleading or fake content, by the time the report was published. The Iranian government also spread misleading footage after the attack, with a state TV broadcast showing videos that it claimed demonstrated a missile’s impact in Israel, but were in actuality footage from wildfires in Chile. X has significantly reduced its trust and safety staff since Elon Musk bought the app in October 2022, with the company firing over 80 percent of its trust and safety engineers between October 2022 and January 2024, according to a report from Australia’s online safety commission.

The United Kingdom Ministry of Justice announces new draft law prohibiting the creation of explicit content

The United Kingdom Ministry of Justice announced a new draft law on Tuesday that criminalizes the creation of sexually explicit deepfakes and pornographic images without a person’s consent. The draft law will be introduced as an amendment to the Criminal Justice Bill, which is currently making its way through Parliament. It is already illegal to share deepfakes of someone without their consent in the UK, but the new amendment would make the UK the first to criminalize the creation of an explicit deepfake—whether it is shared or not. The Minister for Victim and Safeguarding, Laura Farris, stated that the amendment “sends a crystal-clear message that making this material is immoral, often misogynistic, and a crime.”

China Academy of Information and Communications Technology’s (CAICT) releases first results from AI safety benchmarking initiative

The China Academy of Information and Communications Technology (CAICT)—a government-affiliated think tank under China’s Ministry of Industry and Information Technology—worked with seventeen other groups to launch large-model safety benchmarks earlier this week. CAICT evaluated its benchmark criteria against eight models, including those developed by Alibaba and several major Chinese universities, and conducted automated and manual reviews of individual models’ responses to 7,343 test questions separated into three major categories: technology ethics, data security, and content security. The models were given a responsibility and safety score, with responsibility scores determining whether and how well a model answered a question and safety scores determining whether the answer was acceptable. The average safety score amongst the eight models was 93.8, and the average responsibility score was 55.32. China has worked to roll out better evaluation benchmarks for general purpose models, but researchers in the country, and in many others, have faced difficulties in creating benchmarks that are both specific and useful to consumers.

 

Cecilia Marrinan is the intern for the Digital and Cyberspace Policy Program.

More on:

Artificial Intelligence (AI)

Digital Policy

Iran

Israel

Creative Commons
Creative Commons: Some rights reserved.
Close
This work is licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) License.
View License Detail