Anthropic launches Claude Opus 4.7 with intentionally limited capabilities

Claude Opus 4.7 is characterized by notable improvements in practical domains such as software engineering, instruction following, and the execution of complex tasks. The model also benefits from enhanced memory management via file handling, facilitating its integration into professional environments. However, Anthropic emphasizes that it remains overall less effective than Claude Mythos Preview, its most advanced model, specifically regarding cybersecurity skills, which have been deliberately restricted.

This limitation is part of a "security by design" approach. Opus 4.7 is equipped with mechanisms capable of automatically identifying and blocking requests deemed sensitive or prohibited. The company explained that it adjusted the model's training to specifically reduce certain high-risk capabilities, while continuing to draw insights from real-world deployments to prepare future versions that are both more powerful and better controlled.

Finally, Anthropic is offering regulated access to sensitive features for cybersecurity professionals through a dedicated verification program. This strategy illustrates the company's commitment to balancing technological innovation with rigorous usage management, in a context where the most advanced models require reinforced safeguards.