Anthropic Expands Secure AI Access to Three Tiers
AI TIMES ·
✦ AI Summary
According to AI TIMES, Anthropic overhauled its Cybersecurity Vulnerability Program on October 6 and decided to provide AI access rights in…
According to AI TIMES, Anthropic overhauled its Cybersecurity Vulnerability Program on October 6 and decided to provide AI access rights in three tiers depending on the work purposes of security organizations. The move comes as at least 129,000 verified vulnerabilities were identified in projects such as Glasswing, and it aims to widen advanced capabilities needed for defense while addressing the risk of misuse for attacks through verification and controls. The revised plan consolidates the existing program into one framework and reviews defensive organizations, approved penetration-testing organizations, and specialized security organizations that deal with life and critical infrastructure under different standards. It also made clear that while basic security code analysis and patch support will remain available to general users, more sensitive functions will be opened only to organizations whose identity and purpose have been verified. The company said that even with the same model, different safeguards can be applied depending on the level of verification. The measure shows that in cybersecurity AI, purpose-based access control is becoming a core competitive advantage over blanket restrictions.
Perspective
The significance of this issue is that competition in security AI is shifting from the model's performance itself to operational capability: how it is designed to open access to whom, and how far. On the defense side, organizations can use the necessary functions more quickly, while providers look for a balance that reduces liability from indiscriminate access. In the end, the industry is likely to see a stronger trend toward differentiation based on user verification, privilege separation, and the sophistication of control design, even when using the same technology.
This perspective is BizCrush's own commentary and is not part of the reporting by AI TIMES.
This article was produced with the help of an automated content generation algorithm.
Source: AI TIMES
View originalThis article was summarized and organized by BizCrush based on the original article from AI TIMES. For exact quotations and full details, please refer to the original article.