Anthropic, a US-based AI startup, is investigating reports of unauthorized access to its unreleased Claude Mythos model, which is designed to bolster cybersecurity. A Bloomberg report revealed that a small group accessed the model via a third-party vendor environment on the same day Anthropic announced its limited release to select companies, including Goldman Sachs and Apple. Although the users did not run cybersecurity prompts, their access has raised serious concerns about access controls and the potential misuse of the technology. Claude Mythos is a large language model (LLM) capable of identifying zero-day vulnerabilities in software systems and is part of Anthropic's Project Glasswing, which aims to enhance defensive cybersecurity capabilities. The model has drawn scrutiny from the UK’s AI Security Institute due to its advanced capabilities, including the potential to orchestrate multi-step cyber-attacks autonomously. Anthropic has pledged up to $100 million in usage credits and $4 billion toward open-source security initiatives, highlighting its commitment to addressing the challenges posed by sophisticated AI-driven cyber threats.