The digital ecosystem of humanity has been thrown into chaos as Anthropic, a U.S. generative artificial intelligence (AI) developer, recently unveiled its cybersecurity AI "Mithos." Citing concerns over catastrophic consequences if immediately released, the company has made it available on a limited basis to Microsoft, Google, and Apple through "Project Glasswing," a pre-deployment collaboration initiative among Big Tech firms. An AI service evaluated as having the most powerful capabilities in cybersecurity has emerged as a threat to humanity, with the potential to hack the security networks of every financial and government institution worldwide.
AI is penetrating deeply into our lives. With the advent of the generative AI era and the emergence of assistant-type agents, some jobs were predicted to disappear while new ones would be created. Subsequently, the development of autonomous agent AI capable of independent judgment is shaking the labor and industrial structure to its core. Not only is work speed being dramatically shortened, but technical barriers in coding and design work, as well as boundaries between industries, are all breaking down.
AI is also delivering a major shock to the educational field. The social authority of professors, who produce knowledge, and of academic degrees has declined, while classrooms and campuses are losing their presence as spaces of learning. "Super individuals" armed with AI are enhancing their capabilities, evolving through personalized learning and real-time feedback on their own. Moreover, the advancement of blockchain technology is leading a network-based learning paradigm, mutually verifying and recording individuals' lifelong learning experiences.
The advancement of AI particularly threatens the public sphere of democracy. Deepfakes, which are fabricated videos, and the "filter bubble" phenomenon that reinforces confirmation bias can harm citizens' rational thinking and healthy debate, while critical national policy decisions can be distorted by algorithmic bias. There is concern that the formation of public opinion in the public sphere, a core process of democracy, could be distorted by technology.
However, there are also high expectations that AI advancement will lead to fair policymaking and legislation. Italy's "AI-Augmented Parliament" uses AI to analyze the similarities between bills and produces rational and fair alternatives through real-time revisions. The U.K. government's "Consult" uses AI to analyze and organize vast amounts of opinions submitted during the legislative process and then presents the most reasonable policies. Meanwhile, IT firms such as FiscalNote in the United States and CODIT in Korea analyze extensive data on laws, policies, and legislative activities to predict the passage prospects of bills of interest and regulatory directions, providing information to stakeholders. "AI politicians" are also running for office. Although they lack eligibility to be elected, they have run on behalf of humans under the banner of correcting corrupt and unfair politics. Representative AI politicians include Sam in New Zealand, Steve in the U.K., and Robama in the U.S.
For AI to minimize risk factors and become a technology for the human community, not only states and markets but also civil society and the international community must jointly build a governance framework. There is an urgent need to establish a cooperative governance system at the legal, ethical, and technical levels to protect humans from the polarization of the public sphere, social discrimination, and privacy violations brought about by AI advancement.
