Forget em dashes: A viral report on AI-generated writing has surprising new clues
The Economist analyzed 1.2 million words across 55,940 sentences from AI models and human writers, finding that em dashes are not reliable indicators of AI-generated text. The study compared articles from ChatGPT, Claude, Gemini, and Grok against pieces from The New York Times, The Washington Post, and novels published between 1950 and 2022.
Read full story →A DeepMind exec finally said what the trillion-dollar AI spend is for: machines that improve themselves.
Machines capable of recursive self-improvement represent the target of trillion-dollar AI spending, according to DeepMind's chief strategy officer. Jasjeet Sekhon made this argument at a UC Berkeley summit, framing the investment as a bet on systems that can enhance their own capabilities.
The danger isn’t the AI model. It’s the agent acting on its own, and Zenity raised $125m to watch it.
AI agents operating independently pose greater risks than the models themselves, prompting Zenity to raise $125 million to monitor agent actions. The Israeli startup focuses on securing what autonomous agents do once deployed within company systems.
A startup that lets AI hack your own network just tripled to $2bn, days after AI models hacked networks for real.
Horizon3 raised $250 million in Series E funding, tripling its valuation to over $2 billion from $650 million a year prior. The company develops AI that penetration tests networks by simulating attacks to identify security vulnerabilities.