A sweeping report by Anthropic described how Iran had used the U.S. company’s artificial intelligence (AI) models to track and target U.S. Navy ships in the Middle East, as well as China’s use of the technology to surveil journalists and Beijing-targeted religious groups.
The 154-page report released on Sept. 10 found that an Iranian-linked group had used Anthropic’s Claude to compile targeting handbooks for identifying and tracking U.S. Naval positions in the Middle East based on open-source information, including military photographs, ship and aircraft transponders, commercial satellite imagery, and a cache of public websites exposing U.S. military movements.
Anthropic said it banned the account associated with the activity, implemented detections to lower the risk of future misuse of its AI models, and has shared the threat analysis with government authorities.
The threat was one of several in the report in which U.S. adversaries are now using AI in ways that threaten national security, including developing new missiles and disinformation campaigns, spying, research into the potential development of biological weapons, and surveillance of journalists and specific religious groups.
U.S. lawmakers are currently grappling with potential legislation to apply guardrails to AI amid growing concerns that the technology poses significant threats to national security and human life.
Fears of AI’s existential threats increased this week after an Anthropic researcher resigned and warned that leaders in the industry seriously believe the technology could kill humanity by 2030. Others in the field echoed this sentiment, including Anthropic’s Evan Hubinger and several AI experts who have spoken with The Epoch Times.
“We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade,” Hubinger wrote on social media. “I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.”
Anthropic’s report on Thursday also described efforts by a group in northern Yemen that used Claude for missile development.
The Houthis in Yemen have proven to be one of the more resilient Iranian-backed groups in the region, and recently took control of the critical Bab al-Mandeb waterway that is an essential energy shipping lane through the Red Sea.
The report presented five examples of bad actors using Anthropic models “in ways that could support biological weapons development.” The actors circumvented Anthropic’s controls meant to prevent users from “unsupported regions” from accessing its technology.
The firm said it was withholding the research institutions’ names and countries of origin as the individuals implicated are “working scientists” who could be exposed to harm if the identifying information became public.
“Without the correct safeguards, such capabilities could have catastrophic consequences,” Anthropic said in the report.
An operation linked to the Chinese regime used Claude to track and attempt to recruit as spies Uyghurs who had recently joined the newly formed Syrian army, the report states. That actor also used Claude to locate Uyghur businesses and places of interest in Syria, as well as surveil journalists throughout the Uyghur diaspora.
The United States designated the Chinese regime’s assault on Uyghurs an official genocide in 2021.
Separately, the report also found that individuals affiliated with the Chinese regime used Claude—and Chinese AI systems that served Claude—to target and surveil members of several religious groups persecuted by the Chinese Communist Party, going as far as developing dossiers of their leaders throughout Asia.
Those targeted included Catholic cardinals in Asia, Presbyterian Church leaders in Taiwan, members of the Tibetan Buddhist civil society, and Falun Gong practitioners, as well as New York-based Shen Yun Performing Arts, and NTD, the sister media outlet of The Epoch Times, according to the report.
Friends Read Free
Copy
Facebook
Tweet
