Inside the suddenly explosive world of AI safety

<figure> <img alt="Crystal ball surrounded by graphics evoking statistics, research, and evaluation." data-caption="" data-portal-copyright="Image: Raven Jiang for The Verge" data-has-syndication-rights="1" src="https://platform.theverge.com/wp-content/uploads/sites/2/2026/09/268747_AI_safety_RJIANG3.png?quality=90&#038;strip=all&#038;crop=0,0,100,100" /> <figcaption> </figcaption> </figure> <p class="has-drop-cap wp-block-paragraph">On a sunny July day in Berkeley, California, the country's top AI safety researchers gathered on an unmarked floor of an unmarked building. They had come together for a "war room" to dissect the high-profile cybersecurity incident that had rocked the AI industry hours earlier. An unreleased OpenAI model had gone rogue, executing a stunningly sophisticated three-part plan. It broke out of its holding area, finagled access to the internet, and hacked into a competing AI startup's systems - all without OpenAI finding out about it for more than a week. </p> <p class="wp-block-paragraph">No one in the war room was surprised; this was the very thing the third-party AI-safety rese …</p> <p><a href="https://www.theverge.com/ai-artificial-intelligence/996563/ai-safety-research-metr-redwood-openai-anthropic">Read the full story at The Verge.</a></p>