📨 AI Rewrites Climate Science, Job Markets & Cyber Warfare

This week: AI is transforming climate science by improving cloud modeling while simultaneously threatening democratic processes through fake comment campaigns. Additionally, Europe's cybersecurity gap is widening as 80% of phishing attacks now leverage AI.

Clouds represent the largest uncertainty in climate modeling, leading to predictions ranging from 2°C to 6°C warming. Traditional physics models cannot simulate cloud-scale details, which would require 100 billion times current computing power. Researchers are now using AI: Caltech’s CLIMA trains algorithms on over 8,000 simulated clouds to derive better parameters, while Ai2’s ACE2 directly predicts climate from 50 years of real atmospheric data, entirely bypassing physics equations.[QuantaMagazine]

Post-ChatGPT, job postings for structured, repetitive tasks decreased by 13%, while demand for analytical, technical, and creative roles increased by 20%. This is according to Harvard Business School research analyzing US vacancies through March 2025. The largest reductions were observed in the finance and tech sectors. The finding suggests that AI creates demand for augmentation-prone roles requiring human judgment — such as microbiologists, financial analysts, and clinical neuropsychologists — indicating that human-AI collaboration drives labor transformation. [Harvard Business School]

Anthropic’s product releases have caused market turmoil: a legal plug-in led to billions in losses for Thomson Reuters and RELX [Reuters]; a security scanner caused CrowdStrike, Cloudflare, and Okta to drop 8-10% [Reuters]; and a modernization tool resulted in IBM’s steepest decline in over 25 years [YahooFinance]. The core debate centers on whether AI agents automating junior-level tasks threaten per-seat licensing models. While these tools still require human oversight, AI labs are aggressively hiring to bridge enterprise integration gaps.

AI applications demonstrated real-world impact: researchers developed an algorithm that classifies pediatric brain tumors with 92% accuracy from liquid biopsies [STJude], offering a noninvasive diagnostic breakthrough applicable across cancer types.

Regarding adoption, Anthropic launched the AI Fluency Index, which tracks 11 behavioral traits to measure human interaction with AI. This metric is designed to quantify adoption and uptake patterns over time.

Valuation concerns grew as over 1,300 AI startups achieved $100M+ valuations despite questionable fundamentals [Inc.].

Security and trust issues escalated: a small hacker group breached over 600 firewalls using commercial AI tools, demonstrating how AI lowers technical barriers for adversaries [Bloomberg]. Concurrently, AI content authentication tools continue to fail, prompting Microsoft to advocate for new industry standards [MITtechreview].

China’s reported dominance in humanoid robots appears inflated, lacking real commercial demand beyond state-funded showcases [The Economist].

A Miami jury ordered Tesla to pay $240M in damages after determining that “Autopilot” marketing inflated public expectations about the system’s capabilities, contributing to a fatal crash [Reuters].

Capital continues flowing into AI at scale: Meta secures chips, OpenAI raises $100B, and Salesforce accelerates its agentic platform buildout:

AI chip competition intensified as SambaNova raised $350M to challenge Nvidia's hardware dominance [Pitchbook]. Capital flows remained strong: legal AI startups attracted $5B+ in 2025 [Forbes], while the reinforcement learning services sector reached $17B [CNN]. Wall Street is now exploring ways to securitize AI chips and computing power as tradable financial assets amid surging data center spending [The Economist].

US Big Tech AI spending is projected to reach $700B in 2026 (a 75% increase from 2025), exceeding Sweden’s entire GDP. Amazon alone invests $200B, surpassing the combined GDP of the Baltic states. Europe’s sovereign cloud infrastructure totals only €10.6B, making total US spending 66 times larger. Mistral AI’s €1B CapEx and a Swedish data center represent rare European resistance, but critics warn that US “sovereign-light” solutions leave Europe technically dependent on American infrastructure. Wall Street questions sustainability: Morgan Stanley projects $400B in Big Tech debt to fund expansion, double 2025 levels [Euronews].

An AI-powered platform, CiviClick, generated over 20,000 opposition emails that influenced Southern California air regulators to scrap gas appliance phase-out rules aimed at reducing smog pollution. When officials verified a sample, three recipients stated they had not written the messages. Experts warn that AI-generated astroturfing makes it harder for officials to distinguish genuine public opinion, threatening policymaking integrity nationwide [LAtimes].

At the Masters of Digital Forum today in Brussels, EU policymakers and industry leaders emphasized a shift from compliance-based cybersecurity to real capability and readiness. Key concerns included fragmentation across member states, supply chain risks, and AI-enabled threats — 80% of 2025 phishing attacks used AI. The upcoming Cybersecurity Act revision aims to reduce fragmentation. The main conclusion of the EU’s Cyber-Resilience discussion was that the problem lies not in regulation but in ineffective implementation.

👀 Climate Model Revolution's Investment Implications: AI bypassing physics equations could accelerate climate predictions, reshaping risk models for infrastructure, agriculture, and energy investments. Watch for increased compute demand as models shift from simulation to data-driven approaches.

👀 AI's Labor Market Bifurcation Accelerates: Harvard research shows 13% drop in repetitive roles vs. 20% surge in analytical jobs post-ChatGPT. Watch finance and tech sectors for continued hollowing-out of junior positions, an expanding wage premium for augmentation-resistant skills, and potential consumer spending impacts as displacement outpaces job creation velocity.

👀 AI-Powered Cyberattacks Outpace Defenses: Hackers breached over 600 firewalls using commercial AI tools, while 80% of last year’s phishing attacks were AI-enabled. Europe has regulations but lacks effective implementation. Watch for an acceleration in breach frequency as the execution gap widens between policy and practice.

📅 16–17 June 2026 | 🌐 University of Salford, Manchester & Online

he 3rd EPEAI international research conference will focus on how AI is reshaping business models, productivity, labor markets, and economic governance. It will feature dedicated tracks on AI-driven entrepreneurship, finance, HR, supply chains, GenAI and decision-making, regulation, skills, and competitiveness. The conference brings together researchers, policymakers, and practitioners to debate the economic implications of AI diffusion for firms, markets, and society. More info / register

“Invert, always invert” — 19th-cent. mathematician Carl Jacobi developed a fundamental insight: Many math problems become easier when worked backwards. This wisdom transcends fields: doctors pledge “First, do no harm," Michelangelo carved David by removing what wasn't David, Warren Buffett built his track record on 2 rules: (1) Don't lose money, (2) Don't forget Rule 1.
Across centuries and disciplines, one timeless principle.

Which brings us to AI. In a space dominated by hype and FOMO, inversion becomes essential discipline: Instead of asking What will succeed? ask What consistently fails?

That’s it for this week.

Until next time,

The CLNM Capital

[Edited with the assistance of LLMs]

Want to share this newsletter with a friend? Use this link: https://clnm.beehiiv.com/