By Rafael Floqi/
The narrowing artificial-intelligence gap between the United States and China is no longer merely a contest between technology companies. It is becoming a struggle over economic power, national security, military advantage and global influence. As increasingly capable models appear in rapid succession, humanity faces an uncomfortable question: Are we building the most useful instrument in history—or accelerating toward a danger we may be unable to control? The United States remains a leader in advanced models, research and high-performance chips, but China is rapidly closing the distance. Chinese companies such as DeepSeek, Moonshot AI, Z.ai and Alibaba have demonstrated that powerful systems can be developed despite American restrictions on access to cutting-edge semiconductors. According to the Associated Press, experts now describe the American advantage as narrow and potentially fragile.
President Donald Trump summarized the competitive mentality in four words: “Whoever wins AI, wins.” His administration’s AI strategy similarly declares that the United States must prevail in the global race. The White House AI Action Plan presents artificial intelligence as a foundation of economic competitiveness, national security and American technological leadership. China sees the contest in similarly strategic terms. Beijing is pursuing technological self-reliance while warning that AI can be used for cyberattacks, data theft, political manipulation and deepfake propaganda. Both governments therefore recognize the danger, but neither wants to slow down while the other continues racing ahead.
This creates what might be called the AI security trap. If Washington pauses, it fears Beijing will take the lead. If Beijing exercises restraint, it fears permanent dependence on American technology. When both sides believe caution benefits the opponent, speed becomes a measure of survival. Safety testing, transparency and public oversight may then be treated as delays instead of necessities.
The greatest immediate dangers are already visible. AI can produce convincing false photographs, voices and videos within seconds. It can imitate public officials, manipulate voters, generate propaganda and overwhelm social media with fabricated events. It can assist hackers in identifying software vulnerabilities, automate surveillance and make fraud cheaper and more convincing. It may also eliminate or transform millions of jobs before governments, schools and workers have time to adapt.
These threats do not require a conscious or evil machine. They require only powerful systems placed in irresponsible human hands. The longer-term danger is more disturbing. Developers are building “agentic” systems capable of planning and executing complicated tasks with decreasing human supervision. An AI does not need emotions, hatred or a desire for domination to cause catastrophic harm. It needs only a badly defined objective, access to important systems and enough autonomy to pursue its instructions in ways its creators did not anticipate.
The leaders of the AI industry have themselves acknowledged this possibility. Sam Altman of OpenAI, Demis Hassabis of Google DeepMind and Dario Amodei of Anthropic signed a declaration stating: “Mitigating the risk of extinction from AI should be a global priority.” The Center for AI Safety published the statement alongside signatures from hundreds of researchers and public figures.
OpenAI has also warned that superintelligence “could lead to the disempowerment of humanity or even human extinction.” Its statement on controlling superintelligent systems acknowledged that existing methods may not be sufficient to supervise machines significantly more capable than their human overseers. These warnings should be taken seriously, but not transformed into theatrical certainty. There is no evidence that today’s AI systems possess consciousness, independent ambition or a secret intention to destroy humanity. They generate outputs by processing enormous quantities of data and identifying patterns. They can make remarkable discoveries, but they can also fabricate facts, misunderstand instructions and reproduce human prejudice.
Therefore, “Apocalypse Now” is not a description of present reality. It is a warning about a possible future created by political rivalry, corporate competition and weak safeguards. Panic can be as dangerous as complacency. Exaggerated predictions may convince the public that catastrophe is inevitable and that ordinary people have no role in shaping the future. They may also distract attention from harms already occurring: discrimination by automated systems, loss of privacy, deepfake fraud, intellectual-property disputes, concentration of power and the displacement of workers.
At the same time, dismissing every warning as science fiction would be irresponsible. Humanity has repeatedly introduced transformative technologies before fully understanding their consequences. Nuclear weapons created destructive power capable of ending civilization, yet international agreements, monitoring systems and communication between rivals reduced some of the risks. AI requires a comparable seriousness, although its software can spread more easily than nuclear material and is far more difficult to contain.
The United States and China do not need to trust one another completely to recognize shared interests. Neither country benefits from uncontrolled autonomous weapons, AI-assisted biological attacks, attacks on hospitals and electrical grids, or machines making irreversible military decisions without meaningful human authorization. Agreements on testing, incident reporting, cybersecurity and human control over nuclear systems could reduce common dangers without ending technological competition.
AI companies must also accept responsibilities proportional to their power. Independent safety evaluations should occur before exceptionally capable systems are widely released. Developers should report serious failures, protect training systems from theft and explain what their models can and cannot reliably do. Governments, meanwhile, must avoid both extremes: regulation so rigid that it freezes legitimate innovation, and regulation so weak that corporations effectively police themselves.
Artificial intelligence could accelerate medical research, improve education, help scientists understand climate systems and give people powerful new creative tools. It could also strengthen dictatorships, industrialize deception and place weapons or infrastructure under inadequate human control. The technology does not guarantee either future. Human decisions will determine which becomes more likely.
The central question is therefore not whether America or China will win the AI race. It is whether humanity can prevent the race from eliminating every meaningful brake on innovation. The apocalypse is not here. But the warning lights are visible. If intelligence advances faster than wisdom, law and international cooperation, the final failure may not come from a machine that rebels against humanity. It may come from humanity becoming so frightened of losing the race that it forgets why the race was being run.
Foto: https://ekofin.al