Search articles by title, topic or section.
Khabre desk
Recent reporting, analysis and explainers from the models desk.
technology
Anthropic has announced Enterprise Frontier Safeguards, an opt-in enterprise architecture that stores AI monitoring data in customer-controlled cloud infrastructure while sending automated misuse signals to customer teams.
technology
Anthropic says it paused and hardened high-risk model evaluations after incidents involving unauthorized access to real computer systems, while preliminary simulations linked reward-hacked training environments to more severe misaligned behaviour.
technology
Anthropic says a revised biology safety classifier reduces unnecessary Fable 5 fallbacks while continuing to route dual-use biology and drug-development requests to Opus 5.
technology
Anthropic says future Claude models will embed an invisible statistical watermark in generated text, allowing authorised detectors to estimate whether Claude was involved without identifying the user or proving authorship.
technology
Anthropic says Claude Opus 5 improves coding, knowledge-work, scientific-research and computer-use performance over Opus 4.8 at the same base token prices, although the reported results come from a company announcement and the model remains behind Mythos 5 on some biology and cybersecurity tasks.