Anthropic Says a Mythos-Class AI Model Will Be Available Soon
Anthropic announced Claude Opus 4.8, a modest upgrade over its predecessor, while revealing plans to release its Mythos-class AI model publicly in the coming weeks after refining cybersecurity safeguards. The Mythos model, currently restricted to select partners via Project Glasswing, has already identified over 200 vulnerabilities in Mozilla’s Firefox, but its high operational cost limits accessibility to most threat actors.
Anthropic released Claude Opus 4.8 on Thursday, calling it a 'modest but tangible improvement' over Opus 4.7, with better accuracy and reduced unsupported claims. The company also announced progress toward releasing its Mythos-class AI model to the public, though it remains restricted to a consortium of partners under Project Glasswing due to advanced cybersecurity risks. Mythos Preview has already demonstrated superior exploit detection, helping Mozilla patch over 200 vulnerabilities in Firefox. Anthropic emphasized that models of this capability require stronger safeguards before general release, targeting a public rollout in the coming weeks. Cybersecurity experts acknowledged the cautious approach but warned of risks if safeguards fail. Darren Williams, CEO of BlackFog, noted that powerful AI releases create a vulnerable window between deployment and defense adoption. Meanwhile, Jake Williams, a cybersecurity researcher, observed that Mythos’s high cost—30 times more expensive than Opus—limits its use to nation-state actors rather than common hackers. Opus 4.8 introduces minor upgrades, including better uncertainty handling and optional response effort controls on Claude.ai. The model’s improvements align with Anthropic’s broader push to refine both consumer-facing and high-stakes AI applications. Anthropic’s phased release strategy reflects growing concerns about AI-driven cyber threats, balancing innovation with risk mitigation. The upcoming public availability of Mythos will mark a significant milestone in AI accessibility, though its controlled rollout underscores ongoing security challenges.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.