After Mythos, end laissez-faire on AI

TL;DR

America must regulate AI after Anthropic's Mythos model showed unprecedented hacking power.
Five AI leaders—Dario Amodei, Demis Hassabis, Elon Musk, Mark Zuckerberg, Sam Altman—control the tech.
Laissez-faire policy fails politically and strategically against China.

The story at a glance

The Economist's leader argues that the Trump administration's hands-off approach to AI can no longer hold after Anthropic's Mythos model demonstrated extreme capabilities in finding software vulnerabilities. It names Dario (Anthropic), Demis (Google DeepMind), Elon (xAI), Mark (Meta), and Sam (OpenAI) as wielding godlike control over powerful AI. The piece is timed to Mythos's restricted release via Project Glasswing in early April 2026, which alarmed governments and banks.[[1]](https://www.economist.com/leaders/2026/04/16/america-wakes-up-to-ais-dangerous-power?etear=nl_today_2)[[2]](https://www.economist.com/business/2026/04/08/how-dangerous-is-mythos-anthropics-new-ai-model)

Key points

Mythos, Anthropic's frontier model, autonomously found thousands of high-severity bugs in every major OS and browser, including a 27-year-old undetected flaw.[[2]](https://www.economist.com/business/2026/04/08/how-dangerous-is-mythos-anthropics-new-ai-model)
Rather than public release, Anthropic limits access to 40+ firms like Apple, Microsoft, JPMorgan via Project Glasswing to patch vulnerabilities first.[[3]](https://www.economist.com/business/2026/04/15/why-anthropic-and-openai-are-locking-up-their-latest-models)
The five AI bosses command models with jaw-dropping abilities; ChatGPT alone has over 900m weekly users.[[4]](https://www.economist.com/business/2026/04/16/could-ais-leading-men-become-as-powerful-as-ford-or-rockefeller)
Trump officials urged testing Mythos; banks and regulators worldwide met urgently over cyber risks.[[5]](https://thehill.com/policy/technology/5829315-anthropic-mythos-ai-cybersecurity-risks)
OpenAI follows suit, restricting its cyber-focused model, signaling industry shift to gated releases.[[6]](https://www.semafor.com/article/04/15/2026/openai-unveils-a-rival-to-anthropics-mythos)

Details and context

Anthropic announced Mythos on April 7th 2026, pausing broad rollout due to its hacking prowess—far beyond prior models like Claude Opus. It excels at exploits, prompting panic: Fed officials warned bank CEOs, UK and Canadian regulators assessed threats, and Trump aides like VP Vance questioned AI CEOs.[[7]](https://www.theglobeandmail.com/business/economy/article-anthropic-mythos-ai-defence)

The leader critiques laissez-faire as unwise amid US-China rivalry; unchecked private power risks safety and security. Mythos builds on past caution, like OpenAI's 2019 GPT-2 holdback (led by Dario Amodei then). Smaller models match some feats, but Mythos scales uniquely, raising stakes for legacy code.[[8]](https://www.tomshardware.com/tech-industry/artificial-intelligence/anthropics-claude-mythos-might-be-the-best-overall-ai-model-for-cybersecurity-but-cheaper-models-can-attain-similar-results-research-shows-cross-examination-of-the-frontier-model-raises-questions-on-uptime-and-reliability)

Key quotes

"Should a handful of men be entrusted with the world’s most potent new technology?"[[1]](https://www.economist.com/leaders/2026/04/16/america-wakes-up-to-ais-dangerous-power?etear=nl_today_2)

Why it matters

AI's hacking leap threatens global infrastructure, from banks to browsers, amplifying cyber risks in a US-China race. Businesses face urgent patching needs; investors eye gated models boosting firms like Anthropic (valuation bids at $800bn); consumers risk disrupted services.[[6]](https://www.semafor.com/article/04/15/2026/openai-unveils-a-rival-to-anthropics-mythos)

Policymakers must balance innovation and guardrails, as private control grows unaccountable.

Watch regulatory moves from Trump administration and rivals' responses; OpenAI's cyber model rollout could set precedents, though full impacts remain uncertain.[[9]](https://www.cnbc.com/2026/04/16/anthropic-claude-opus-4-7-model-mythos.html)