March2026
Artificial intelligence isn’t a future risk — it’s a present one. In just the past year, AI has rocketed from the 10th to the 2nd most critical global business risk according to the Allianz Risk Barometer, and incidents tracked by the AI Incident Database have already surpassed all of 2024’s totals before 2025 even ended. The technology is moving faster than our ability to govern it, and the consequences are showing up everywhere — in our jobs, our security, our health systems, and even our power grid.
Here’s what the data actually says.
1. Job Displacement Is No Longer Theoretical
The labor market disruption that analysts warned about is now measurable. AI was a significant contributing factor in nearly 55,000 U.S. layoffs in 2025 alone, and the first two months of 2026 have already seen 32,000 tech job losses. Major companies are leading the charge: Amazon cut 15,000 roles, Salesforce eliminated 4,000 customer support positions (citing AI handling 50% of the workload), and Workday slashed 1,750 jobs to redirect resources toward AI investment.
The World Economic Forum’s 2025 Future of Jobs Report projects 92 million jobs displaced globally by 2030 — though it also forecasts 170 million new roles created, yielding a net gain of 78 million. The problem is the transition. The jobs being eliminated and the jobs being created require vastly different skills, and the people losing work aren’t necessarily the ones who can fill the new roles.
Worker anxiety reflects this. Employee concern about AI-driven job loss jumped from 28% in 2024 to 40% in 2026, according to Mercer’s Global Talent Trends survey. Workers aged 18–24 are 129% more likely to worry about obsolescence than those over 65. Sectors facing the steepest exposure include customer service (80% of roles projected to be automated), legal support (paralegals face an 80% automation risk), and data entry and administrative work (7.5 million jobs potentially eliminated by 2027).
The displacement isn’t gender-neutral either. Roughly 79% of employed women in the U.S. work in jobs at high risk of automation, compared to 58% of men. Among the 6.1 million workers Brookings identifies as lacking the adaptive capacity to manage transitions, 86% are women.
2. Deepfakes Have Become a Default Attack Vector
The numbers here are staggering. Deepfake incidents surged from roughly 500,000 in 2023 to an estimated 8 million in 2025 — annual growth approaching 900%. Since 2023, deepfake video incidents have outnumbered those from autonomous vehicles, facial recognition, and content moderation algorithms combined.
The financial toll is mounting. Deepfake fraud caused over $200 million in losses in North America in Q1 2025 alone. AI-facilitated fraud losses in the U.S. are projected to reach $40 billion by 2027, growing at a 32% compound annual rate. In one high-profile case, a finance worker at engineering firm Arup was tricked into wiring $25 million after a deepfake video conference call convincingly impersonated company executives.
What makes 2025 a turning point is the emergence of Deepfake-as-a-Service (DaaS) platforms — ready-made tools that put voice cloning, video synthesis, and persona simulation into the hands of anyone willing to pay. According to Cyble’s threat monitoring data, AI-powered deepfakes were involved in over 30% of high-impact corporate impersonation attacks in 2025. Voice cloning has crossed what researchers call the “indistinguishable threshold”: a few seconds of audio now produces a convincing clone.
A 2025 iProov study found that only 0.1% of participants correctly identified all fake and real media presented to them. Humans can no longer reliably tell what’s real. Gartner predicts that by 2026, 30% of enterprises will no longer trust standalone identity verification solutions.
3. AI-Powered Cybercrime Is Scaling Fast
Malicious use of AI for scams and disinformation has grown 8-fold since 2022. Underground marketplaces now sell pre-packaged AI tools that dramatically lower the skill threshold for launching cyberattacks — you no longer need to be a hacker to hack. In 2025, an AI agent placed in the top 5% of teams in a major cybersecurity competition, demonstrating that these systems can discover and exploit software vulnerabilities at expert level.
AI is being used to generate phishing emails that are grammatically flawless and contextually personalized, to write malicious code, and to automate the discovery of zero-day vulnerabilities. The 2026 International AI Safety Report warns that as more organizations deploy AI systems, the attack surface expands — not just through the models themselves, but through the complex infrastructure built around them.
4. Algorithmic Bias Is Causing Real-World Harm
AI bias isn’t an abstract concern — it’s producing measurable discrimination in healthcare, hiring, and everyday interactions.
In healthcare, medical AI systems have contributed to a 30% higher mortality rate for non-Hispanic Black patients compared to white patients. A 2025 Cedars-Sinai study found that leading large language models generate less effective treatment recommendations when a patient is identified as African American.
In hiring, the picture is equally troubling. In 2026 testing, AI hiring tools selected Black male names 0% of the time in resume screening. A federal judge in 2025 allowed a collective action lawsuit to proceed against Workday, alleging its AI screening tools disproportionately disadvantaged applicants over 40. Stanford research published in Nature found that models like ChatGPT carry deep-seated biases against older women in the workplace.
Even image generation systems encode harmful stereotypes — a 2025 study found that braids and natural Black hairstyles consistently received lower “intelligence” and “professionalism” scores from AI evaluation tools. These aren’t edge cases. They’re systematic patterns baked into the training data and architectures that millions of people interact with daily.
5. The Environmental Cost Is Enormous and Growing
The energy appetite of AI is reshaping power grids. U.S. data centers consumed 183 terawatt-hours in 2024 — more than 4% of the nation’s total electricity, roughly equivalent to Pakistan’s entire annual demand. Global data center electricity consumption is projected to reach 1,050 TWh by 2026, with the IEA projecting 15% annual growth through 2030.
The carbon and water footprints are substantial. AI systems alone may have produced between 32.6 and 79.7 million tons of CO2 in 2025, with water consumption reaching up to 764.6 billion liters — comparable to the world’s entire annual bottled water production. Approximately 60% of data center energy still comes from fossil fuels.
The strain on infrastructure is already creating real problems. PJM Interconnection, the largest U.S. grid operator serving 65 million people across 13 states, projects it will be six gigawatts short of reliability requirements by 2027. In some regions, data center demand is already pushing residential electricity bills up by $16–18 per month. In Ireland, data centers could consume 32% of the country’s electricity by 2026.
6. Biological and Weapons Risks Are Emerging
This is perhaps the most sobering category. In 2025, multiple AI companies released models with heightened safeguards after pre-deployment testing could not rule out that their systems could meaningfully help novices develop biological weapons. The 2026 International AI Safety Report flagged this as a growing concern, noting that the capability bar continues to drop as models become more powerful.
AI systems can now synthesize knowledge across domains in ways that lower the expertise required to cause catastrophic harm. The safeguards in place are largely voluntary — only 12 companies published or updated Frontier AI Safety Frameworks in 2025, and most risk management remains self-regulated.
7. Governance Can’t Keep Up
The overarching theme across all of these risks is a governance gap. AI capabilities are advancing faster than the rules, institutions, and safety measures designed to manage them. The 2026 International AI Safety Report — a collaboration involving researchers from 30 countries — concluded that global risk management frameworks remain immature, with limited quantitative benchmarks and significant evidence gaps.
A few jurisdictions are beginning to formalize requirements. South Korea enacted a comprehensive AI Framework Act effective January 2026, mandating fairness and non-discrimination in high-impact sectors. The EU AI Act is being phased in. But most of the world still relies on voluntary commitments from the companies building these systems.
Meanwhile, 23% of companies are already scaling AI agents within their organizations and another 39% are experimenting, according to McKinsey. As CNBC reported in March 2026, the most pressing risks may not come from the models themselves but from the complex, poorly understood systems that companies build around them — creating what experts call “silent failure at scale.”
What This Means
None of this is to say AI doesn’t offer tremendous potential. It does. But the data makes clear that we’re in a period where the harms are accelerating alongside the benefits, and our institutions haven’t caught up. The risks are not hypothetical — they’re measured in job losses, fraud dollars, discriminatory outcomes, carbon emissions, and emerging threats to biosecurity.
The question isn’t whether AI is dangerous. The data already answers that. The question is whether we’ll build the governance, safeguards, and transition support systems fast enough to keep the damage manageable.
Sources: Allianz Risk Barometer 2026, AI Incident Database, World Economic Forum Future of Jobs Report 2025, Mercer Global Talent Trends 2026, Brookings Institution, Cyble Executive Threat Monitoring, iProov 2025, Gartner, International Energy Agency, PJM Interconnection, International AI Safety Report 2026, McKinsey State of AI 2025, Deloitte Center for Financial Services, Cedars-Sinai Medical Center, Stanford University/Nature.
