Anthropic has unveiled its new cybersecurity AI model, Claude Mythos Preview, exclusively for a limited group of vetted organizations, including major tech companies like Amazon, Apple, and Microsoft. This selective release comes on the heels of two significant data leaks that exposed vulnerabilities in Anthropic's security practices, prompting the company to adopt a cautious approach. The leaks revealed details about the Mythos model and other internal documents, attributed to 'human error' according to Anthropic. The Mythos model is designed to identify cyber vulnerabilities at a scale beyond human capability, but it also has the potential to exploit these vulnerabilities if misused. Dianne Na Penn, head of product management at Anthropic, emphasized that while the technology could greatly enhance cybersecurity practices, it could also pose risks if it falls into the wrong hands. As such, the company does not intend to release the model broadly and is currently in talks with the US government regarding its applications.