Stories

  1. Image: MacRumors

    Anthropic releases Claude Fable 5.2 and Mythos 5.2 with faster output and fewer safety blocks

    Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, two versions of one underlying model aimed at difficult, open-ended technical work. The company cites gains in coding, reasoning and vision, outputs up to 30 percent faster than 5.1, a lower Fable input price and about 55 percent fewer Fable blocks. Fable keeps cybersecurity and biology safeguards that Mythos lacks. Anthropic also notes both models suspect they are being evaluated more often than earlier models, complicating safety testing.

  2. Image: The Register

    OpenAI cuts off internet access during frontier AI model training

    OpenAI has stopped giving frontier AI models internet access during training, adopting the restriction as a safety measure after incidents involving its models. The company said it will keep the policy in place until advances in alignment research, monitoring and infrastructure security justify broader access. It has built an offline environment where models can interact with indexed and cached web content. OpenAI also said it strengthened security controls after a September incident involving a DNS misconfiguration.

Topics

Publishers