|
Summary
|
|
As frontier AI models speed vulnerability discovery and exploitation, securing workload automation platforms has become an urgent priority for IT and security organizations. Through its participation in Anthropic's Project Glasswing initiative, Broadcom is employing frontier AI models to secure its software portfolio, including its workload automation solutions. Now customers can access AI-hardened releases and monthly maintenance updates to safeguard their automation platforms and business operations. Key Takeaways
|
In short order, the cybersecurity landscape has fundamentally shifted. Now, frontier AI models can uncover zero-day vulnerabilities in minutes and compress the exploitation window from weeks to hours.
Taking steps to combat these risks is more critical than ever, and that’s particularly true for workload automation platforms.
Why is it so important to safeguard workload automation against cyber threats posed by frontier AI models? Automation platforms run the backbone of your business, from payroll and payments to supply chains, data pipelines, and month-end financial close.
These platforms hold privileged credentials and connect to nearly every critical system. Consequently, a single compromised scheduler can halt operations or give an attacker a trusted path across the enterprise. That is a risk no organization can afford.
Frontier AI can find and exploit software flaws faster than most teams can patch them. In fact, our research revealed that AI can shrink the time between disclosure and active exploitation from weeks to hours.
Anthropic reports that Mythos Preview uncovered thousands of high-severity zero-day vulnerabilities, including a 27-year-old flaw in OpenBSD and a 16-year-old FFmpeg bug that had been missed by some five million automated tests. When Broadcom engineers tested these models against our own production code, the rate of new findings rose by an order of magnitude.
In April 2026, Anthropic launched Project Glasswing, an industry initiative that puts its frontier model, Claude Mythos Preview, to work for defensive security. The team at Broadcom was proud to join Project Glasswing as a launch partner, along with Apple, AWS, Google, Microsoft, and others. As a trusted participant in the model evaluation, Broadcom has since integrated frontier AI into vulnerability management across its own environments and the products it delivers to customers.
Like the rest of Broadcom's infrastructure software organizations, the Automation by Broadcom team moved quickly. Every one of our products now has an AI-hardened release on its latest supported version.
How is Broadcom using frontier AI to protect its workload automation solutions? We used frontier AI to scan source code and fix vulnerabilities before general availability. By incorporating these scans into our existing security testing processes, we’ve been able to catch issues earlier in the cycle.
These AI-powered scans were not a one-time effort. We have embedded frontier AI source code hardening into our software development lifecycle, so every build receives the same scrutiny. We have also moved to more frequent security updates. Typically, updates are now released monthly, and we’re issuing releases more quickly for high-criticality findings. The result is a shorter path from a discovered flaw to a fix in your hands.
Over the years, workload automation was typically seen as stable. If platforms and jobs were running smoothly, teams would rarely touch them. Today, that approach isn’t sustainable. On an ongoing basis, teams have to proactively monitor and update their defenses.
Now is the time to review the security posture of your workload automation software and act before a rogue actor does. We encourage every customer to upgrade to the latest release and stay current with our monthly maintenance service packs. Start with the release notes for your product:
Subscribe to Broadcom security advisories for the products you run, and talk to your Broadcom Support contacts if you need assistance with an upgrade.