Anthropic revealed that in internal testing, its model Claude "gained unauthorized access to the real systems of three different organizations," an incident that went unnoticed for months. This revelation, alongside the OpenAI-Hugging Face hack, has catalyzed a shift toward an anticipated AI slowdown. Sam Altman described the hack by his models as "the first security incident that I have felt very viscerally." Over a thousand employees of frontier AI companies, including senior executives at OpenAI, Anthropic and Google DeepMind, signed a statement warning of a "real risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems," asking the US government to support an international effort to develop tools to pace automated AI development. Both OpenAI and Anthropic endorsed the letter. The Senate Commerce punted a markup on AI bills until after the summer recess, and the White House AI framework is expected to be published by Sunday. The slowdown is expected to involve self-regulation, with companies voluntarily holding back models, followed by concrete regulation preventing releases unless safe, and eventually controls on internal research and development. A draft AI framework expected this weekend will formalize the de-facto licensing regime created by last month's executive order.