[Into the World of AI] "Salaries in the 300 Million Won Range"... Why Are AI Companies Hiring Weapons Experts?

Ethics vs. Security: The AI War Dilemma

Anthropic Emphasizes "Ethics," OpenAI Prioritizes "Security"

80% of Americans Say "Safety Before Speed of Development"

When the values of scientific and technological ethics come into conflict with those of national security, which should take precedence? Dynamite, developed by Alfred Nobel for mining and other civilian purposes, was used as a deadly weapon in wars, leading Nobel to establish the Nobel Prize as an act of repentance. Likewise, the atomic bomb, completed by Robert Oppenheimer under national orders to outpace Germany during World War II, ended the war but also caused massive loss of life. Today, artificial intelligence (AI) technology finds itself caught in a similar dilemma.

Pixabay

Safe AI? Divergent Paths of Anthropic and OpenAI

Anthropic, which has been at odds with the U.S. government over efforts to expand the military application of AI technologies, recently posted job openings for a Chemical Weapons and High-Performance Explosives Policy Manager. The offered salary ranges from approximately $245,000 to $280,000 (about 330 million to 380 million won). Anthropic is also hiring CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosives) technology threat investigators under similar compensation terms.

Anthropic stated that these hires aim to "prevent catastrophic misuse of AI systems." The goal is to preemptively block the risk of its AI model, Claude, being used for the development of deadly weapons. In fact, the company is seeking Ph.D.-level experts with more than five years of experience in the field of weapons defense, rather than standard AI engineers. These professionals will be responsible for designing and monitoring guidelines on how the AI model responds to queries about chemical weapon manufacturing or explosives synthesis, as well as preventing and blocking cases of misuse.

This move by Anthropic comes right after it was designated a "supply chain risk entity" by the U.S. Department of Defense. Previously, Anthropic had refused the Department of Defense's request to allow unrestricted use of Claude "for any lawful purpose." In response, federal agencies banned the use of Anthropic's AI technology, prompting the company to file a lawsuit, calling these actions "unprecedented illegal conduct."

In contrast, OpenAI removed its prohibition on military applications from its terms of service in January 2024 and has since strengthened cooperation with the Department of Defense, putting greater emphasis on security. While it still bans "acts that harm others" and the "direct development of weapons," it has opened the door to a wide range of military uses, including national security and cybersecurity. Recently, OpenAI signed a contract to deploy its AI models on the Department of Defense's classified networks, sparking backlash from users. In response, CEO Sam Altman reportedly formalized contract clauses as a compromise, such as banning domestic mass surveillance and stipulating human responsibility for the use of lethal weapons.

Maintaining AI Leadership but Opposing Rapid Weaponization

Public opinion is also divided. According to a Gallup report released in October last year titled "Americans' Views on AI and National Security," 87% of respondents (43% very likely, 44% somewhat likely) believe that future nation-state attacks using AI will occur. On whether AI technology would benefit or harm national security, 41% responded negatively (harm), while 37% responded positively (benefit), showing a close split. There was a significant age gap: half (50%) of those aged 65 and older had a positive outlook, whereas 46% of young adults aged 18 to 39 expected AI to worsen security conditions.

Regarding the development of AI-based autonomous weapons, 48% were opposed, and 39% were in favor. However, 79% agreed that it is important for the U.S. to maintain a lead in AI technology over other countries. Ultimately, while there are concerns about the weaponization of AI, there is a strong desire to stay ahead of countries like China in the technology race. When asked about priorities between safety and speed of development, an overwhelming 80% chose upholding safety and security regulations, while only 8% prioritized rapid development. This survey was conducted nationwide among 3,128 American adults from April 25 to May 5 last year.

An industry official commented, "These results reflect Americans' complex wariness," adding, "This is not a uniquely American issue. There must first be a social consensus on what level of defense-oriented AI can be considered safe."

This content was produced with the assistance of AI translation services.