Safely fixing cloud vulnerabilities requires understanding how they are exploited, but that kind of research runs into guardrails on most AI platforms.
Legitimate defensive security work shares DNA with offensive techniques. Exploitability analysis, adversarial testing, root-cause investigation all require reasoning about how an attacker would operate.
Frontier AI models apply safeguards that, by default, limit this dual-use work. For general use, those safeguards make sense. For security teams building remediation intelligence, they create a real barrier.
Anthropic built the Cyber Verification Program (CVP) to address this. And Tamnoon has been accepted.
Verified access will strengthen the research behind Tamnoon’s cloud remediation platform and help organizations turn CNAPP findings into closed issues with greater confidence.
Why Joining the Cyber Verification Program Matters Now
Frontier AI models are increasingly central to how security teams research, analyze, and respond to threats. But the same capabilities that help defenders model attack paths and assess exploitability can be misused by bad actors.
Anthropic has addressed this from two directions:
- Project Glasswing gives critical infrastructure organizations access to Claude Mythos Preview for finding and fixing vulnerabilities in foundational software.
- The Cyber Verification Program takes a different angle, letting vetted cybersecurity organizations conduct legitimate dual-use defensive research on Claude’s Opus models with fewer default interruptions.
Both programs share an underlying premise. As AI capabilities accelerate, defenders need access to the same reasoning power that attackers will inevitably reach. The question is whether that access arrives through structured, responsible channels.
What the Cyber Verification Program Is
The CVP is a free, application-based program from Anthropic designed for cybersecurity organizations whose legitimate defensive work overlaps with dual-use techniques.
Accepted organizations can work on tasks, like vulnerability research, exploitability analysis, and adversarial simulation, with fewer of the default safeguards that would otherwise interrupt this work. Access is scoped to defensive use cases and governed by Anthropic’s security, confidentiality, and responsible-use requirements.
The program is separate from Project Glasswing, which provides access to Anthropic’s restricted Mythos model. CVP applies to Claude’s Opus models and focuses on removing friction from legitimate security research workflows.
How Tamnoon Uses CVP Access
Tamnoon’s agent-led, expert-supervised remediation model depends on understanding how vulnerabilities behave in real cloud environments.
Tami goes beyond closing tickets. It investigates findings, assesses remediation safety using the Remediation Confidence Indicator (RCI), and creates plans to fix the root cause. That process requires deep knowledge of how misconfigurations and vulnerabilities actually get exploited in production.
Here’s how CVP access strengthens our research:
- Deeper exploitability analysis: Understanding whether a given vulnerability can actually be exploited in a specific cloud environment is what separates a noisy finding from a genuine risk. Verified access lets Tamnoon’s team reason through these scenarios more thoroughly, improving the accuracy of RCI scores and prioritization decisions.
- More rigorous remediation validation: Every remediation plan Tami generates needs to close the exposure without introducing new risk. CVP access lets the team apply adversarial thinking to Tamnoon’s own playbooks, stress-testing proposed fixes against the same logic an attacker would use.
- Stronger root-cause investigation: Many cloud security issues recur because fixes address symptoms instead of causes. Modeling how attackers chain findings together helps Tamnoon identify the underlying misconfiguration or architectural pattern that produces the alert in the first place.
The output of this research lands in the platform customers already use. There’s no separate product or new interfaces. Simply put, the intelligence behind Tami gets even better.
What This Means for Tamnoon Customers
Nothing changes about how customers interact with Tamnoon. The platform, the workflow, the agent-led, expert-supervised model all remain the same. What improves is the depth and accuracy of the research powering it.
Sharper exploitability data feeds into more accurate RCI scores. Better root-cause models mean fewer recurring alerts. More thoroughly validated remediation playbooks mean safer outcomes in production.
Tamnoon already delivers up to 80% reduction in open exposures within 90 days. CVP access strengthens the research foundation that makes those results possible, with the same headcount and without disrupting operations.
Stronger Research, Safer Remediation
Tamnoon’s mission is closing the last mile of cloud security, the gap between detection and resolution. That work depends on understanding vulnerabilities as deeply as the people who exploit them.
Acceptance into Anthropic’s Cyber Verification Program is one step in building the research foundation that supports safe, accurate, production-ready AI-driven remediation at scale. We’ll share more as the program develops.
