While 89% of security leaders say they're prepared for AI-driven attacks, most organizations still rely on manual vulnerability management and human intervention, as Kai introduces Auto Remediation to help close the gap

SAN JOSE, Calif. - July 29, 2026 - Kai, the company behind the first agentic AI cybersecurity platform designed to execute security work end-to-end at machine speed with human expert accuracy, today announced findings from its inaugural 2026 State of Autonomous Defense Report, and the availability of its Auto Remediation capability.  

According to the global survey of 500 CISOs, organizations face a growing mismatch between machine-speed attacks and human-speed response. As AI continues to reshape the cyber threat landscape, security leaders increasingly recognize that traditional, human-led approaches to vulnerability management are no longer sufficient. Kai’s report found that while CISOs overwhelmingly see machine-led security as the future, barriers including governance and operational readiness still exist.

"AI is transforming how organizations defend themselves and shifting the economics of cyber defense," said Galina Antova, co-founder and CEO, Kai. "Speed alone isn't enough. The real opportunity today is trusted automation that doesn't just prioritize risk, it eliminates it, giving security teams a clear path to zero. That's why we built Auto Remediation: to close the loop between detection and action. The next phase of cybersecurity must let security teams respond at the pace modern threats demand."

CISOs understand the AI threat, but many security processes remain human-led

Security leaders overwhelmingly acknowledge that AI has changed the threat landscape. Nearly nine in ten respondents (89%) say their organization is prepared for AI-driven attacks, yet fewer than one-third (28%) describe themselves as very prepared, underscoring the gap between awareness and confidence. Additionally, 63% believe attackers currently have the advantage because of AI. This comes amid the rapidly shrinking time-to-exploit (TTE) across the industry, demonstrating the severity of how fast attackers find and exploit vulnerabilities.  

As attackers increasingly leverage AI to accelerate exploitation, many organizations continue to rely on manual security workflows that struggle to keep pace.

Remediation can’t keep pace with AI

The research found that vulnerability management remains heavily dependent on manual effort, leaving organizations exposed and security teams overwhelmed. Sixty-five percent of organizations report their vulnerability management processes are at least half manual, while 60% require more than one week to remediate critical vulnerabilities. Nearly half (48%) say at least one-quarter of known vulnerabilities go unremediated for more than 30 days.

These operational challenges are taking a measurable toll on security teams. More than three-quarters (77%) of respondents say vulnerability management contributes at least moderately to burnout, and 60% or more report that remediation validation and automated actions still require human intervention.

The findings suggest it's not just the threat landscape creating risk, but the operating model itself.

Organizations want more automation, but barriers remain

While organizations increasingly see automation as essential to keeping pace with AI-powered threats, confidence in autonomous decision-making has yet to catch up. More than half (52%) of respondents identify lack of trust in automated decisions as the biggest barrier to broader automation adoption, followed by governance and compliance concerns (43%) and skills and talent gaps (38%).

Organizations are already embracing automation for lower-risk activities such as asset discovery and inventory (55%) and vulnerability prioritization (49%). That comfort is extending further, with 32% of organizations already permitting automated remediation actions, a meaningful signal that CISOs are growing comfortable letting machines change the environment when the context is right, and increasingly trusting automation to make the call.

Advanced organizations are already accelerating when it comes to machine-led security

Despite today's challenges, the research points to a clear shift toward more autonomous security operations. Today, 35% of organizations describe their vulnerability management approach as mostly or primarily machine-led. Over the next 12 to 18 months, that figure is expected to increase to 45%, signaling growing confidence that automation will become foundational to future cyber defense.  

With Auto Remediation from Kai, every confirmed finding gets a decision: Is this a real risk, and if it is, can Kai fix it without a human in the loop, or does it require human intervention? If it can, Kai remediates it automatically. If not, Kai builds the full remediation plan and stages the change so technical teams can act in minutes instead of days. It's automation that draws the line where organizations need it today, with room to extend that trust as confidence grows. Read more about Auto Remediation on the Kai blog.

Read the full 2026 State of Autonomous Defense Report, here.  

Methodology

The Kai Survey was conducted by Wakefield Research (www.wakefieldresearch.com) among 500 CISOs at private sector companies with a minimum annual revenue of $500m. The research was conducted in four markets between June 15th and June 29th, 2026, using an email invitation and an online survey.

About Kai

Kai is the AI company rebuilding cybersecurity for the machine-speed era. Trusted by Fortune 500 and Global 2000 enterprises, the Kai Autonomous Defense Platform replaces fragmented tools and human-limited workflows with agentic AI that works continuously across cyber asset management, application security, infrastructure vulnerability management, and detection engineering. It contextualizes, reasons, and acts at machine speed and enterprise scale. What takes human-led teams weeks, Kai executes in hours, driving risk toward zero through Auto Remediation. Human defenders don't just keep up. They become superhuman.

‍