Author: lethuphuong

📌 The International Monetary Fund (IMF) believes that the global economy in 2026 will still grow stably at 3.3% despite trade risks, thanks to strong momentum from AI investment and business adaptability. AI acts as a growth catalyst; if adopted quickly and effectively, global growth in 2026 could increase by an additional 0.3 percentage points. In the medium term, AI could contribute an additional 0.1–0.8 percentage points annually to global growth. The trend of falling inflation creates a monetary policy foundation to support growth in the coming period.

📌 The ASEAN region currently has over 680 AI companies, with about 500 (over 70%) located in Singapore. The region accounts for only about 2% of total global AI funding, lower than its 4% share of global GDP. Total venture capital investment in Southeast Asian AI startups in 2025 is estimated at $410.5 million, down from $520.2 million in 2024. Facing AI bubble risks, ASEAN startups are strategically shifting towards profitability and real cash flow. Lean enterprises focused on specialized niches and real paying customers are seen as candidates to overcome any upcoming market corrections.

📌 Current AI systems often lack consistency: objects deform, spaces change, and time gets “broken” in videos or simulations. The core cause is that generative AI operates on a mechanism of probabilistic prediction, without maintaining a continuous world model to update its understanding. World models are emerging as the foundation for the next wave of AI, solving AI’s biggest current weakness: a lack of stable understanding of space and time. From video, AR, and robotics to AGI, the ability to build and update continuous world models could determine whether AI merely “mimics” or truly understands and acts correctly in the…

📌 Doctors are beginning to record cases where users exhibit psychotic symptoms, in which AI – especially chatbots – plays a central role. “AI psychosis” is not an official medical diagnosis, but a term describing psychotic symptoms shaped or amplified by interaction with AI. Current AI safety mechanisms focus mainly on self-harm and violence, not yet focusing on psychosis. Although not directly causing psychosis, highly interactive AI can amplify delusions and blur the boundaries of reality. The solution is not to boycott AI, but to closely combine technological design, medicine, and ethics to protect mental health in the AI era.

📌 The UK government has appointed two “AI Champions” to lead the deployment of AI in financial services, aiming to boost growth, productivity, and investment while protecting consumers and maintaining system stability. The roles take effect from January 20, 2026, are direct ministerial appointments, and are unpaid. With 75% of financial businesses already using AI and a potential to create an additional $25–38 billion USD in value by 2030, the UK is shifting from encouragement to strategic coordination. The appointment of AI Champions demonstrates an effort to balance accelerated innovation with risk governance. If successful, this will be a lever…

📌 AI is not a single bubble but multiple stacked bubbles. The AI ecosystem is divided into 3 layers: wrapper companies, foundation models, and infrastructure. Wrapper companies are at risk of collapse within 18 months, foundation model companies will consolidate within 2–4 years, while infrastructure companies, despite short-term excess, will retain long-term value. Advice for builders: the biggest risk is not building a wrapper, but remaining just a wrapper; one needs to own the workflow, data, and distribution channels.

📌 ChatGPT Free currently meets basic needs and light usage well, while Plus is the most balanced choice for frequent users wanting fewer restrictions and early access to GPT-5.2, Codex, Agent, and Sora for $20. Pro is only truly reasonable for research experts, programmers, or businesses needing maximum power, as the $200 price tag mainly trades for high limits, fast speed, and the GPT-5.2 Pro model.

📌 Mistral’s CEO and co-founder argues that the company’s biggest competitive advantage in Europe comes not from smarter AI models, but from being a non-US option that aligns with the need for control and technological sovereignty. With a $14 billion valuation, military contracts for the French Ministry of Defense, and an open-source strategy, Mistral exploits the demand for sovereignty, control, and trust from governments and regulated enterprises. Their long-term advantage is not just the model, but their geographical location and their approach to building AI.

📌 Canada has never officially defined “sovereign AI,” allowing the concept to form loosely through data center funding, domestic cloud partnerships, and data residency requirements. Consequently, “sovereignty” is applied inconsistently, relying on infrastructure location rather than true control. Businesses optimize for rapid deployment on foreign platforms to meet geographic requirements while lacking technical and legal control. For example, despite government investment in Cohere, the compute infrastructure is operated by US-based CoreWeave, subject to US law. If Canada wants to build a sustainable AI ecosystem, sovereignty standards need to shift from “where” to “who has the power,” so that value, knowledge,…

📌 Experts urge Malaysia to soon issue clear regulations to manage foreign AI platforms like DeepSeek, aiming to protect data privacy and prevent external influence. Malaysia faces a strategic choice regarding foreign AI like DeepSeek: prohibition or smart governance. Experts agree that risks regarding data, bias, and political influence are real, but the sustainable solution is a clear legal framework, risk assessment, and the development of domestic AI. With the AI law expected to be presented in June 2026, how Malaysia balances innovation and data sovereignty will determine the safety and reliability of the national AI ecosystem.