Ethical Hacking News
Anthropic has expanded its access to advanced AI models, providing vetted cybersecurity professionals with a new Cyber Verification Program (CVP). This move comes on the heels of the discovery of over 129,000 verified software vulnerabilities and aims to empower organizations and security teams to secure systems using the same capabilities that could be weaponized by bad actors. With its three access tiers, the CVP is designed to provide a safe and controlled environment for security teams to test and deploy AI models, addressing the growing concern of AI-powered zero-day attacks.
Anthropic launches a new Cyber Verification Program (CVP) to provide vetted cybersecurity professionals with access to its AI models. The CVP aims to empower organizations and security teams to leverage Anthropic's AI models in their defensive efforts. The program features three access tiers: Defense Access, Red Team Access, and Specialized Access, catering to different needs and requirements. The CVP has been influenced by the growing dual-use nature of AI capabilities, aiming to balance security and access. The program is designed to address the growing concern of AI-powered zero-day attacks and provide security teams with tools to fight back.
In a move that marks a significant expansion of its Project Glasswing initiative, Anthropic has announced the launch of a new Cyber Verification Program (CVP), designed to provide vetted cybersecurity professionals with access to the company's advanced artificial intelligence (AI) models. This development comes on the heels of the discovery of over 129,000 verified software vulnerabilities between April and July 2026, as well as an additional 5,500 vulnerabilities uncovered through open-source scanning efforts between April and October 2026.
The CVP, which features three access tiers, aims to empower organizations and security teams to leverage Anthropic's AI models in their defensive efforts, while also facilitating the testing of these models with reduced safeguards and blocking classifiers. This initiative is seen as a step towards making advanced cyber capabilities safely available to a broader set of defenders, in line with Anthropic's dual-use nature.
The company's Project Glasswing initiative, which was launched in 2025, has been instrumental in uncovering a substantial number of software vulnerabilities. According to a recent analysis, only 2 of the 300 vulnerabilities discovered by Anthropic or Project Glasswing have been exploited in the wild. Furthermore, a survey of verified vulnerabilities found that over 33,000 have been rated as critical- or high-severity, with the true impact likely to be at least five times higher.
The CVP's three access tiers – Defense Access, Red Team Access, and Specialized Access – cater to different needs and requirements. The Defense Access tier is designed for defensive work, such as incident response, malware reverse engineering, and vulnerability analysis and validation. The Red Team Access tier adds authorized penetration testing and red-teaming to the defensive use cases, while the Specialized Access tier has the fewest safeguards and is reserved for a limited set of verified organizations that are authorized to test safety systems.
According to a CyScenarioBench evaluation, the safeguards implemented in the CVP allowed 46 of 50 tasks on Claude Opus 5.5 in the Defense Access tier to be blocked, while the Red Team Access tier did not block any tasks, completing 34 of 50 tasks at the same completion rate as when no safeguards were applied. This evaluation provides confidence that Anthropic's AI models can be made available safely to a broader set of defenders.
The decision to expand access to its AI models has been influenced by the growing dual-use nature of these capabilities. On one hand, these models can be used to secure systems, while on the other hand, they can be weaponized by bad actors for malicious purposes. By providing vetted cybersecurity professionals with access to its AI models, Anthropic aims to empower security teams to secure systems using the same capabilities that could be used to exploit them.
The company's initiative is also seen as a response to the growing concern of AI-powered zero-day attacks. According to a recent report, attackers are using AI to accelerate reconnaissance, compromise identities, and escalate access. Anthropic's CVP is designed to address this concern by providing security teams with the tools and capabilities needed to fight back against these types of attacks.
In an effort to provide more visibility, controls, and governance around AI agents operating within enterprises, Anthropic has also launched a number of resources and tools. These include a practical guide to defending identity against AI-powered attacks, a video series on securing, investigating, and deploying AI, and a SANS training program for app sec and cloud teams who inherited GenAI risk.
The launch of Anthropic's CVP marks a significant expansion of its Project Glasswing initiative, which has been instrumental in uncovering a substantial number of software vulnerabilities. As the company continues to push the boundaries of AI research and development, it is clear that its commitment to making advanced cyber capabilities safely available to a broader set of defenders is a key priority.
Related Information:
https://www.ethicalhackingnews.com/articles/Anthropic-Expands-Access-to-Advanced-AI-Models-Amidst-Discovery-of-Over-129000-Software-Vulnerabilities-ehn.shtml
https://thehackernews.com/2026/10/anthropic-expands-claude-access-for.html
Published: Wed Oct 7 05:19:17 2026 by llama3.2 3B Q4_K_M
