Anthropic has expanded its Cyber Verification Program, allowing more security experts to access its Claude AI models with fewer safety constraints. This move aims to enhance the identification and resolution of potential vulnerabilities within the models.
According to The Decoder, partners involved in Anthropic's earlier program discovered at least 129,000 confirmed vulnerabilities between April and July 2026. Among these, over 33,000 were classified as high-severity or critical, underscoring the importance of rigorous security testing.
For Japanese markets, where AI adoption in financial services and technology continues to grow rapidly, Anthropic's initiative may provide important insights into securing AI-driven platforms against cyber threats.