The Cybersecurity and Infrastructure Security Agency (CISA) on April 16 added a previously patched high-severity Apache ActiveMQ remote code execution (RCE) bug to its Known Exploited Vulnerabilities (KEV) catalog.Security experts thought adding the bug to the KEV made sense because Apache ActiveMQ is one of the most popular open-source Java-based message brokers with thousands of deployments worldwide and has long been a favorite target of attackers.Another important reason for adding it to the KEV was that the vulnerability — CVE-2026-34197 — went undetected for 13 years until it was discovered by Horizon 3ai researcher Naveen Sunkavally, who posted April 7 that he found it using the Claude AI assistant.It’s important to note that Sunkavally used the publicly available Claude AI assistant, and not Claude Mythos, which has mostly been limited to the 40 organizations involved with Project Glasswing coalition of tech companies using the AI model to scan for zero-day bugs that has created a stir in the industry since its introduction in April.Sunkavally, a distinguished engineer at Horizon 3ai, said while much has been made of Mythos, we are already at the inflection point where AI models are significantly accelerating vulnerability research. Sunkavally said publicly available models like Opus 4.6 are already highly capable in the hands of experienced researchers.“Mythos and future models will likely make it possible for less experienced researchers or even laypeople to find vulnerabilities with minimal guidance,” said Sunkavally.
As for the ActiveMQ RCE bug being added to the KEV, Sunkavally said his team sees ActiveMQ all the time in enterprise environments, and it has a history of being targeted by threat actors.“While this vulnerability requires credentials to exploit, ActiveMQ is commonly configured with default credentials and in a few versions no credentials are required at all, effectively turning this into an unauthenticated RCE vulnerability,” explained Sunkavally.Noelle Murata, chief operating officer at Xcape Inc., said adding a 13-year old Apache ActiveMQ flaw to CISA’s KEV catalog marks a pivotal moment where AI’s "commoditized" discovery has collided with decades of technical debt.Murata said while ActiveMQ’s massive footprint makes it a target, its KEV status was driven by a dangerous "capability leap."“Using a standard Claude AI assistant, researchers condensed a week’s worth of manual analysis into 10 minutes, uncovering an RCE vector that had eluded human eyes since 2012,” said Murata. “The reality is that we’ve been writing mediocre, unvetted code for decades, and our industry has largely relied on ‘security through obscurity’ by assuming no one would bother to read it. But the robots have arrived, and they are exceptionally good at the one thing humans find miserable: meticulously auditing millions of lines of legacy garbage.”Murata added that the real danger isn't a failure of the ActiveMQ architecture: it's that attackers no longer need elite researchers to weaponize 13-year-old mistakes. They just need a subscription and a prompt.“Security teams must prioritize disabling the Jolokia interface /api/jolokia/ or updating immediately, because the ‘patch your stuff’ mantra has now evolved into a race against an adversary that never gets bored and reads code faster than our developers can defend it,” said Murata. “If it takes an LLM 10 minutes to find a bug we’ve missed for 13 years, maybe we should spend less time worrying about ‘AI safety’ and more time worrying about the humans.”
As for the ActiveMQ RCE bug being added to the KEV, Sunkavally said his team sees ActiveMQ all the time in enterprise environments, and it has a history of being targeted by threat actors.“While this vulnerability requires credentials to exploit, ActiveMQ is commonly configured with default credentials and in a few versions no credentials are required at all, effectively turning this into an unauthenticated RCE vulnerability,” explained Sunkavally.Noelle Murata, chief operating officer at Xcape Inc., said adding a 13-year old Apache ActiveMQ flaw to CISA’s KEV catalog marks a pivotal moment where AI’s "commoditized" discovery has collided with decades of technical debt.Murata said while ActiveMQ’s massive footprint makes it a target, its KEV status was driven by a dangerous "capability leap."“Using a standard Claude AI assistant, researchers condensed a week’s worth of manual analysis into 10 minutes, uncovering an RCE vector that had eluded human eyes since 2012,” said Murata. “The reality is that we’ve been writing mediocre, unvetted code for decades, and our industry has largely relied on ‘security through obscurity’ by assuming no one would bother to read it. But the robots have arrived, and they are exceptionally good at the one thing humans find miserable: meticulously auditing millions of lines of legacy garbage.”Murata added that the real danger isn't a failure of the ActiveMQ architecture: it's that attackers no longer need elite researchers to weaponize 13-year-old mistakes. They just need a subscription and a prompt.“Security teams must prioritize disabling the Jolokia interface /api/jolokia/ or updating immediately, because the ‘patch your stuff’ mantra has now evolved into a race against an adversary that never gets bored and reads code faster than our developers can defend it,” said Murata. “If it takes an LLM 10 minutes to find a bug we’ve missed for 13 years, maybe we should spend less time worrying about ‘AI safety’ and more time worrying about the humans.”
