The Race to Fix: How Dell Is Using Advanced AI to Stay Ahead of Emerging Threats

Key takeaways 4 min read

• As part of Anthropic's Project Glasswing, Dell Technologies is using Claude Mythos Preview to proactively find and fix vulnerabilities in its products at a sophistication and speed that traditional methods cannot match.
• Frontier AI models have reached a capability threshold that presents the potential to shift the longstanding imbalance between defenders and attackers. Seizing that advantage requires organizations to move with urgency.
• Organizations need to reduce the attack surface, apply security updates quickly and build cyber resilience for AI-fueled threats.

For decades there's been an imbalance in cybersecurity between defenders and attackers. Frontier AI models have reached a new level of capability with the potential to shift that imbalance in the defenders' favor. But seizing that advantage requires moving with urgency and using an underutilized advantage defenders have always had: complete control of their digital terrain.

As part of Anthropic's Project Glasswing, Dell is using Claude Mythos Preview to aggressively find and fix vulnerabilities in our products and newly developed code at a sophistication and speed that traditional methods cannot match. This strengthens our secure product development and vulnerability response practices, helping our customers stay resilient as threats are expected to grow exponentially in speed and scale. We are also applying these advanced AI models to enhance our cybersecurity program to strengthen the security of our internal systems and networks. This bolsters our longstanding commitment to protecting our customers' data and privacy.

A shift in the threat landscape

Software has always contained flaws requiring finding and fixing, but frontier AI models like Mythos Preview, OpenAI's GPT-5.5-Cyber and others have changed the equation. They can find security vulnerabilities at a speed and scale that wasn't previously possible and automate the process of fixing or exploiting them, depending on the operator's motives. While these models are currently reserved for vetted defenders, similar capability is almost certain to become available to threat actors in the months ahead. As a result, organizations and governments are now in a race against attackers to find and fix vulnerabilities first, and speed is the decisive factor.

We are committed to responsibly using Mythos Preview and other advanced AI capabilities as we continue our mission to build the world's most secure and resilient technology. Our goal is to find and remediate vulnerabilities at speed and disclose available fixes so customers can protect their environments before threats materialize.

The race underway between defenders and attackers is one of the most consequential security challenges the industry has ever faced. That same urgency must apply to every organization managing technology in this environment.

What organizations should do now

The urgency of this moment is real. The fundamentals of cybersecurity and resilience haven't changed, but the margin for action is smaller than it's ever been and the consequences have never been greater.

Speed is the variable that matters most right now. The window to get ahead of AI-fueled threats is open, and organizations that move quickly will be far better positioned for the impending tsunami of vulnerabilities and attacks on the horizon. Here is where organizations should focus their attention:

• Reduce exposure. Maintain a current software and vendor inventory. Stay on supported versions, apply updates quickly and follow security configuration guidance. Understand where third-party components and dependencies create exposure. As vulnerability remediation accelerates across the industry, the ability to triage and deploy updates in hours, not weeks, is itself a core component of cyber resilience.
• Use AI to keep pace. Manual vulnerability management workflows can no longer match the volume and speed of today's threat environment. Incorporate AI tools for vulnerability detection, prioritization and remediation to ensure defenses move as fast as the threats. Publicly available frontier models are incredibly capable at finding previously undiscovered vulnerabilities and supporting remediation activities.
• Strengthen cyber resilience. As AI-fueled threats could grow exponentially in the coming months, recovery readiness becomes as important as prevention. Ensure business continuity and recovery plans account for worst-case scenarios, including major Internet platform outages and complete organizational compromise. Define the minimum viable business processes needed to keep operating during catastrophe-level disruption. Deploy an isolated cyber recovery vault for the most critical data. Establish a crisis management construct to help navigate organizational decision making in the face of unparalleled digital disruption.

The power of industry collaboration

Programs like Project Glasswing highlight that no single organization can solve this alone. Industry collaboration and information sharing across technology companies, governments, security researchers and infrastructure providers are essential to ensuring defenders maintain pace - a need increasingly reflected in policy, including the Trump administration's AI Executive Order on securing critical systems against AI-fueled threats. Dell remains committed to collaborating with defenders worldwide as we collectively navigate this unprecedented moment to protect ourselves and our customers.