Everyone's getting hacked

2026-05-13 · Source: Matthew Berman · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Intermediate, extended

Summary

AI-powered cyber threats are rapidly escalating, with Google's threat intelligence group detecting the first AI-discovered zero-day exploit and the widespread "Shy Hallude" npm worm. This surge is attributed to AI accelerating code generation, increasing attack surface, and enabling easier development of sophisticated hacking tools and polymorphic malware. Major incidents, such as the Vercel security breach in April 2026, have been explicitly linked to AI acceleration. While frontier models like GPT 5.5 Cyber and Anthropic's Mythos are being developed for cyber defense, often with strong guardrails, open-source AI models, despite being less powerful, enable malicious actors to conduct a higher volume of lower-value attacks, making them economically viable. The prevailing defense strategy posits that superior AI models, backed by national resources, will ultimately prevail in defense against weaker, less resourced adversarial AI.

Key takeaway

For CTOs and security leaders evaluating their cyber defense posture, the increasing sophistication and volume of AI-powered attacks necessitate a proactive "AI vs. AI" strategy. You should prioritize investing in advanced AI-driven security solutions and hardening critical systems with frontier models, recognizing that while state-level actors may wield the most powerful AI, open-source models enable a long tail of economically viable, high-volume attacks against less protected targets. Consider implementing robust supply chain security and employee training against AI-enhanced social engineering.

Key insights

AI is rapidly accelerating both cyber attack capabilities and defensive measures, creating an "AI vs. AI" security paradigm.

Principles

• AI accelerates vulnerability discovery, not creation.
• Economic resources dictate AI model superiority in cyber defense.
• Open-source AI lowers the barrier for widespread, lower-value attacks.

Method

Google's GTIG tracks AI-enabled operations, from vulnerability exploitation to augmented operations and initial access, using insights from Mandiant and Gemini to highlight AI's dual role as both an attack engine and a target.

In practice

• Implement a code word with family to counter deepfake phishing.
• Harden software using advanced AI models like GPT 5.5 Cyber.
• Be wary of open-source AI's potential for malicious fine-tuning.

Topics

• AI-Powered Cyber Attacks
• Zero-Day Exploits
• Supply Chain Attacks
• AI in Cybersecurity
• Frontier AI Models

Best for: CTO, VP of Engineering/Data, Investor, AI Security Engineer, Director of AI/ML, Policy Maker

Related on AIssential

• this is really bad... — AI significantly amplifies cyberattack capabilities, necessitating advanced AI-driven defenses and highlighting geopolitical risks.
• IBM and OpenAI Bring Frontier AI to Cyber Defense—Helping Enterprises Keep Pace with Machine-Speed Threats — IBM and OpenAI are integrating frontier AI into cyber defense to counter machine-speed threats and enhance vulnerability management.
• AI-powered hacking has exploded into industrial-scale threat, Google says — AI-powered hacking is an industrial-scale threat, with commercial models used by state and criminal actors.
• Once, cyber-attacks required great skill. AI is changing that | Bruce Schneier — AI democratizes sophisticated cyber-attacks, making robust AI-driven defense an urgent necessity.
• Opinion | America’s Reactive AI Defense Isn’t Enough — AI-driven discovery collapses the time between cyber reconnaissance and exploitation, straining traditional defenses.
• Frontier AI and the Future of Defense: Your Top Questions Answered — Frontier AI models accelerate vulnerability exploitation and social engineering, demanding machine-speed defense and proactive security measures.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Matthew Berman.