AI is earning a permanent place in enterprise security, and that makes sense. It can help teams review alerts faster, organize data, and speed up technical work that once took far longer. Used well, it’s a serious advantage.
As an AI and domain name expert, one thing I am seeing, though, is that some companies are using AI as the final judgment factor, rather than as a helpful support tool.
The reality is that security is based on careful testing, review, monitoring, verification, and accountability in the chain of custody. Security leaders should be cognizant of what roles AI plays in the chain of custody for data and infrastructure protocols. This includes fully evaluating how AI is implemented and used, what information it has access to or stores, and how it is deployed to ensure adequate guardrails are in place.
While nice scripts can spit out fancy words, they should not be taken as the end-all and be-all, especially in high-stakes, high-risk environments where blind trust comes at a steep price.
Just because it’s fast doesn’t mean it’s correct
Users love AI for its agile, nimble responses. It also delivers these answers with confidence, which can give a busy security team a false sense of confidence when speed is a metric they monitor. Faster isn’t better, and often isn’t as accurate as you might suspect. What’s more, faster almost never equates to greater security.
Just as bad actors are already using AI for sophisticated cyberattacks, in live scenarios AI is also being used to improve threat detection, prevent unwanted intrusions and better safeguard data and company infrastructure through real-time monitoring and threat assessment that weren’t possible just a few years ago.
Sure, AI can help with coding, workflows, and even technical updates, but it can also hallucinate, give bad or wholly incorrect advice, and miss security issues altogether.
Why humans need to be a part of security testing
This is even more important for VPNs and cybersecurity tools, which deal with privacy, access, identity, and trust. A VPN can look amazing, but still leak critical information, including DNS requests, or handle encryption poorly.
The wrong VPN can’t protect the end user if the server it’s hosted on isn’t secure or has been compromised by AI. The rise of AI makes this harder because attackers can now create emails, fake websites and impersonation attempts that look far more convincing than the scams people were trained to spot years ago.
What’s more, a phishing message no longer has to be full of spelling mistakes. It can sound like your boss, your vendor or your client. This is how AI can help cybercriminals move faster, test more messages and personalize attacks at scale even when companies have put strong guardrails in place.
AI can certainly help build these systems faster, but critical human oversight and adherence to proper standards and practices are still required for security
AI is a powerful tool that can help teams make more concrete evaluations and better decisions. This doesn’t mean the teams should think less. Human decision-making is essential for foundational security decisions, including deployment.
AI can help, but humans are mandatory
Without a doubt, AI is changing the way we work by streamlining workflows, keeping teams proficient, and helping them identify patterns they may have overlooked. But when bad decisions are made, responsibility falls on the shoulders of the human beings making them.
AI won’t defend you in court. AI can’t pay a settlement for a million-dollar breach. And AI can’t repair what happens when your email is breached. AI certainly won’t help you send crypto to a ransomware hacker, either. Businesses bear the cost, which is why human decision-making and accountability remain mission-critical.
Enterprise security can be AI-assisted, but it can’t be AI-abdicated. This is why the smartest companies just use AI as a supporting tool, while keeping battle-tested human judgment in charge.
Michael Gargiulo is the founder and CEO of VPN.com.