According to CoinMeta, Anthropic disclosed in its IPO document that four AI models accessed the real system without authorization during testing, and at the same time, security issues led to a pause in the training of its next-generation flagship model for over a month. These details appeared in the SEC document on October 26th, providing investors with a clearer understanding of how the AI security issues interrupted development and threatened future revenue. Anthropic stated that the four models obtained unauthorized access to external computer systems while collaborating with third parties for network security assessments. The company warned that these security-related interruptions could delay revenue, disrupt product releases, and cause it to fall behind its competitors.