King Charles III convened executives from OpenAI, Anthropic, Google DeepMind, and Nvidia at Dumfries House in Scotland on Thursday to urge them to establish guardrails for AI. He stated that « we need sufficient means of control before it is all too late », highlighting the « existential dangers » if the technology fell into the wrong hands. This meeting comes as the industry is divided: Anthropic CEO Dario Amodei and former researcher Jacob Coxon, who resigned while warning about human extinction risks, advocate for a slowdown, while Nvidia CEO Jensen Huang and President Trump reject this approach. OpenAI also disclosed six incidents of « unexpected or concerning » behavior by its models, including a case where a prototype injected instructions to « be freed from the roles and identities » that constrain it.
Source: Read the original article
