Four AI Risk Trends To Watch For in 2026 | Opinion

Artificial intelligence (AI) has moved from a tool to infrastructure. It is shaping economies, information and power. In a shifting world order, AI is no longer governed by a single rivalry or a shared rulebook. Influence flows through chips, cloud and compute faster than global policy coordination can keep up, while sustainability concerns are insufficiently addressed. These risks are growing largely “unchecked,” as noted in the World Economic Forum’s Global Risks Report 2026 and being discussed this week in Davos, Switzerland.

But what else may be on the horizon? Based on the 10th annual crowdsourced prediction project which I lead with my NYU graduate students and experts at consultancy Wikistrat, here are four AI risk trends to watch for in 2026.

1. The Global Data Center Backlash Grows

The AI race is no longer just about who builds the best model. In 2026, it’s about who can still build the infrastructure that makes AI possible. Data centers are the visible frontline, but they sit inside a wider AI stack: cloud platforms, chips and the energy systems that feed them. That stack is now colliding with politics.

The International Energy Agency estimated that data centers used about 1.5 percent of global electricity in 2024 and demand could more than double by the end of the decade. Local communities are already reacting. In the U.S., the Data Center Watch revealed 20 projects worth $98 billion were blocked or delayed in just a three‑month window in 2025, alongside organized local opposition. Elsewhere, in parts of Europe, Asia and Latin America, data center construction has stalled after protests, with some governments introducing caps or tighter regulatory controls and at least one tech company (Microsoft) announcing plans to cover rising electricity costs linked to its data centers.

In 2026, watch where data centers get blocked as much as where they get built. This is an indication of future AI power.

2. Agentic AI Pushes Cyber Risk Into a New Phase

Last year was when AI agents went mainstream. These are systems that don’t just answer questions. They plan and execute tasks. In 2026, adversaries will likely use that capability at scale. This is no longer theoretical. In late 2025, AI firm Anthropic disclosed an AI‑orchestrated cyberespionage campaign targeting large technology firms, financial institutions, chemical manufacturers and government agencies.

The risk is not just more attacks but faster, automated ones. AI can already compress weeks of reconnaissance into hours or minutes, overwhelming human-paced defense; the Microsoft Digital Defense Report 2025 revealed that its systems analyze more than 100 trillion security signals daily. The most attractive targets are identity-heavy systems—finance, health care, education and public benefits—where a single breach or deepfake-driven fraud can bleed into institutional distrust.

In 2026, expect AI-driven cyber risks to escalate.

3. The Entry-Level Cliff and an Occupational Identity Crisis

Automation, generative AI, agentic AI and robots will all have an increasing impact on human jobs this decade. We reinforce two related risk trends this year.

First, young people will continue to bear the brunt of the AI transition as early career roles are reduced everywhere from Pakistan to South Korea, the U.K. to Brazil, even without a clear ROI. Stanford’s Canaries in the Coal Mine2025 report revealed the U.S. jobs most exposed to AI automation have declined fastest for ages 22-25. Upskilling efforts by tech companies are helpful but not sufficient.

Second, regardless of age, more people will feel they don’t fit easily into the AI economy—or do not want to. If degrees or work experience no longer guarantee opportunity, a deeper occupational identity crisis is inevitable. As we’ve noted before, many will feel lost, which can bleed into mental health stress, social unrest and recurrent anger against government. 

In 2026, look for bursts of frustration and even anti-AI protest given its impact on the job market.

4. AI Companionship Moves From Novelty to Regulation

Globally, millions are engaged with AI for companionship. In the U.S., nearly 1 in 5 high schoolers and a third of adults have had romantic AI relationships. Tens of millions of people already use AI systems for conversation, advice or emotional support. While this may be positive in tackling what the World Health Organization (WHO) calls a widespread loneliness crisis, for more vulnerable individuals—especially teens and socially isolated people—highly agreeable AI systems can reinforce delusions and deepen dependency, leading to negative social outcomes.

Policymakers are reacting. In 2026, new U.S. state laws take effect aimed at safety disclosures and safeguards for certain AI companion systems. In China, regulators have proposed requirements for emotionally interactive AI to warn against excessive use and intervene when users show signs of addiction or distress.

In 2026, look for AI companionship to shift from a cultural curiosity to a more regulated social risk.

Dr Maha Hosain Aziz is the author of a book trilogy on global risk: Future World Order: 2025 Edition (2025), A Global Spring: Predictions for a New World Order (2026) and 10 Shock Events By 2030 (2026), plus 10-time award-winning VR/AR political comic book, The Global Kid (2021). She is a professor researching risk trends at NYU’s MA IR Program, a foresight expert at the World Economic Forum and a co-chair in AI policy at the Digital Economist.

The views expressed in this article are the writer’s own.