Anthropic releases 'safe' version of Claude Mythos AI model to public (www.theguardian.com)

Anthropic has announced the public release of Fable 5, the first widely available model from its advanced Mythos class, which focuses on safe deployment amid significant cybersecurity concerns. Designed for tasks such as writing and debugging code, answering complex research queries, and analyzing images, Fable 5 represents a strategic shift for Anthropic after months of restricted access due to the model's capabilities in identifying vulnerabilities within critical infrastructures. The company is also offering an unrestricted version, Claude Mythos 5, to select partners, including around 200 organizations in various countries involved in its Project Glasswing cybersecurity initiative. The significance of this launch is underscored by increased scrutiny around AI safety and cybersecurity, especially given the potential threats identified by Anthropic. The firm’s proactive measures, including extensive red-teaming efforts and a bug bounty program, have helped maintain security integrity. With Fable 5 priced at $10 per million input tokens and $50 for output tokens, this release highlights Anthropic's financial strategies amidst ongoing pressure for profitability. Concurrently, the broader AI landscape is heating up with competing firms like OpenAI filing for IPOs, indicating an accelerating race in AI innovation and investment.