AI Outperforms ER Doctors, Fuels Mass Layoffs, and Becomes a National Security Asset in Rapid-Fire Week of Developments

AI is reshaping medicine, the workforce, and national security simultaneously: OpenAI's o1-preview outperforms ER doctors in triage accuracy, Cloudflare cuts 1,100 jobs citing a 600% surge in AI agent usage, and Anthropic refuses a Chinese think tank's request for access to its most restricted model, raising alarms at the White House.

• A Harvard study reveals OpenAI's o1-preview model outperforms two attending ER physicians across 76 real patient cases, achieving a 67.1% triage accuracy versus 55.3% and 50.0% for human doctors, with blind reviewers unable to distinguish AI diagnoses from human ones.
• Cloudflare eliminates 1,100 jobs — roughly 20% of its workforce — directly citing a 600% surge in internal AI agent usage over three months, with agents now capable of autonomously opening accounts, starting subscriptions, and registering domains without human approval.
• A Chinese think tank representative privately urged Anthropic to hand over access to Claude Mythos Preview, its most restricted AI model, at a Singapore meeting — Anthropic refused, prompting alarm from White House National Security Council officials who view the model as a national security asset limited to just 40 U.S. organizations.