When 2025 closes, the overall picture of artificial intelligence (AI) appears with two contrasting colors. The excitement that once enveloped the early stages of AI explosion gradually gives way to a cautious mentality, close supervision and thorny questions about the real limits of this technology.
One of the biggest surprises is the increasingly close relationship between technology corporations and US President Donald Trump.
The US government has eased many barriers related to AI chip exports, promoted the construction of data centers and adopted a lighter regulatory approach. This move helps the AI industry accelerate, but also raises concerns about legal loopholes.
At the user level, 2025 witnessed a strong shift away from traditional search engines. More and more people are using AI chatbots such as ChatGPT or Perplexity to look up information.
Apple's leadership's admission of Safari's search volume decreasing for the first time in more than two decades shows that AI is directly impacting this empire.
The financial market also shook when DeepSeek-R1, an open source reasoning model from China, was launched. DeepSeek's high performance and low cost caused investors to panic, leading to a sell-off of AI stocks, in which NVIDIA lost hundreds of billions of USD in capitalization in just one session. This event shows that the gap between the US and China is narrowing faster than expected.
However, NVIDIA is still a symbol of the AI fever in 2025 when it achieved a record capitalization. However, this overheated increase has raised concerns about the technology bubble, when infrastructure spending is increasing sharply but the long-term profit roadmap is not clear.
In the opposite direction, many failures have cooled down expectations about AI. The incidents related to xAI's Grok chatbot with offensive statements, or ChatGPT being criticized for being sometimes too flattering, sometimes cold, show that the problem of AI behavior control still has no complete solution.
More worryingly, the trend of human-shaped chatbots and personal wearable AI devices raises many questions of ethics, especially for children and privacy. AI agents expected to automate work have also not met expectations, when operations are unstable and legal conflicts easily arise.
Overall, 2025 is not only the year of stronger AI, but also the year the world begins to see more clearly both the potential and the limits of artificial intelligence. From here, the AI race may slow down in speed, but will be stricter in quality, safety and social responsibility.
