“Companies must treat personal data exposure as part of their security risk. This means identifying exactly what information is available, reducing any unnecessary exposure, constantly monitoring for threats, and establishing clear processes for responding when a threat emerges.”
Ron, you founded Ironwall in 2011 around a risk many organizations still treated primarily as a privacy issue. What did you recognize early about the connection between exposed personal information and real-world harm, and how has that understanding evolved over the years?
When we started the company, the consensus in our industry was that personal information was really marketing information and that companies were collecting too much of it. Our realization was that personal information is a lot like a loaded gun; how it is used depends on who wields it. Because we started out protecting some of the most targeted people in an organization, our goal was to keep them safe. So we consulted and hired people who had experience targeting individuals (think special ops) and asked them, “How do you use information to go after someone?” Their answers were eye-opening and gave us the direction that personal information was a vector for attack, and we needed to neutralize that vector.
Your home address, phone number, family info and employment history might seem harmless, but when somebody puts them together, they can create an incredibly detailed picture of your life. This enables them to attack or harass you, threaten your family, target you with sophisticated phishing emails, or even use your info to illegally access your workplace network. One of the sentences we say a lot is: if someone had a grudge against you or your company and knew where you lived, how and when you drive to work, where you drop off your children, and when your family is alone, would that concern you? Because just the most basic of personal information tells that story.
Things are particularly dangerous right now, because AI has made it much easier for even novice fraudsters or laymen attackers to personalize attacks and carry them out at scale.
Nisos’s 2026 executive exposure research found that 94% of executives assessed had home addresses publicly linked to their names. What does this reveal about the gap between an organization’s security posture and the exposure its leaders carry into their personal lives?
Organizations need to change the way they think about security. They invest plenty of money in protecting their networks, facilities and systems, but the information linking their executives to their homes and families is often available to anyone willing to look for it. In fact, it is the combination of networks being generally more secure, along with how much information is out there on individuals, that makes PII-based attacks more prevalent today. It is easier to attack individuals and then use them to gain access to corporate assets, than to go the other way around.
A corporate attacker doesn’t necessarily need to break through the company’s technical defenses. He can start by homing in on a single employee, learning about that person’s family, building a convincing pretext, and using that relationship to gain trust or access.
Executive protection and cybersecurity must be connected. An executive’s personal exposure isn’t a private matter; it can easily become an organizational vulnerability, and companies need to account for it in their security programs.
Also Read: AITech365 Interview with Raj Koneru, CEO/founder of Kore.ai
Ironwall focuses on interrupting the pathway to attacks during the research phase. When personal information is scattered across hundreds of sources, how do you decide what matters most, and what tells you that removing it has genuinely reduced risk rather than simply improved a metric?
Removing one listing isn’t really going to change your risk. What you should be concerned about is what an attacker could learn about you by connecting all the available information that is out there.
We look at the broader exposure picture. Can somebody connect an individual to a home address? Identify their family members? Find a direct phone number? Determine where they work or establish enough personal context to make a fraudulent message believable?
If you want to narrow that down, a home address, a cell phone number and a relative’s name are the three most important pieces of information to find and eradicate. Because with those three pieces of information, you can find out virtually anything else.
Risk reduction is about making this process more difficult. If somebody wants to target an executive but can’t easily uncover their home address or personal contact information, they have to move on to somebody else. That’s what you want, and that’s what we provide. And it isn’t about just reducing the amount of information; you need to focus and remove the key anchor pieces of information and make them extremely difficult to find. When it comes to attacks, time is your friend. The harder you make your data to find, the more likely the would-be attacker moves on, exposes themselves or gets talked out of it by cooler heads.
But the goal isn’t to make it take 10 minutes instead of three to find an executive’s home address, which is the average today. We need to make it take days or weeks. Ideally, we want to force attackers to move beyond online research and into the physical world, where they might have to follow someone home from work. That raises the risk of getting caught or exposing themselves. The longer and harder the research process becomes, the more likely an attacker is to give up and move on to another target.
And removing your personal data from the Internet can’t be a one-time exercise. New data sources constantly emerge and put your information right back up there without your approval. You have to keep monitoring and reducing your exposure on an ongoing basis, rather than treating it as a box to check and forget about, and that means limiting the sources of information as much as removing what is out there. Prevention is as important as removal.
As AI moves from generating answers to using tools, retrieving information, and taking action, security risks are becoming more complex. From your perspective, what changes when an AI system moves beyond assisting a user and starts making decisions or acting on their behalf?
When AI is answering a question, you mainly have to worry about whether the information is accurate. But when it starts using tools, accessing systems, and taking action on its own, that can get very dangerous, very quickly.
For example, an agent with access to sensitive information could share it with other people while initiating actions on behalf of a user. This creates opportunities for attackers to exploit the agent’s access.
Agents can even cause harm without an attacker breaking into the system. For instance, if the agent is given misleading information or manipulated into taking an action, the result can be a severe security incident. What if an AI agent doesn’t treat a password or token as something important? The default set-up for some open source AI agents puts sensitive information in clear text formats. Metaphorically speaking, we are almost back to putting our passwords on sticky notes.
Anybody who wants to use agentic AI needs to think long and hard about what the agents should be able to access, not to mention how to prevent them from taking actions that we don’t want to happen.
An AI agent can encounter malicious instructions hidden in a webpage, email, or document it is asked to process. How should organizations rethink trust when a potential attack can come through the information an agent reads, rather than directly through the system itself?
We’ve seen how publicly available personal information can be used to make a message appear to come from somebody you know or trust. With AI agents, a similar problem arises when the agent reads a webpage, email or document that contains instructions designed to manipulate its behavior.
Just because an agent can read something doesn’t mean it should trust it. Organizations need to test their agents before they are put into production, to make sure they don’t act on malicious instructions, their permissions are limited appropriately, and they will reach out to a human for additional verification before taking any potentially dangerous actions.
This is easier said than done. But in the long run, I feel it will be easier to learn how to manipulate agents than it is to manipulate people. We know how social engineering works on people, and we have had hundreds (or thousands) of years of experience with that. But AI agents are new. Research has shown that you can hack AI agents into giving you restricted information by asking them the same question multiple times. Artificial intelligence does not automatically mean common sense. So for now, I would not scan your credit card into an AI agent or give an AI agent access to network or personal assets.
There is a delicate balance between giving AI agents enough autonomy to be useful and limiting what they can do when something goes wrong. Where should organizations draw that line, particularly when agents can access sensitive information or act across business systems?
I think organizations need to start by asking what could happen if the agent makes a mistake or is manipulated. If an agent is just summarizing info or helping an employee find something, a mistake isn’t that big a deal. But if it can access sensitive records, transfer money, change permissions or communicate externally on someone’s behalf, an error could be devastating.
Never give an agent more access than it needs to do its job, and always establish clear guidelines concerning what it can do on its own and what requires human approval.
Also, plan for failure. An agent could easily get tricked by malicious information or behave in an unexpected way. As such, organizations need the ability to monitor what is happening, a way to stop an agent when something goes wrong, and controls in place that limit the possible damage.
Finally, what if the agent is manipulated by design? If you trust an agent to buy the best airline ticket, or find the best deal, what happens when companies trying to return their investments on AI manipulate the agent into going to a sponsor first, or omitting information that you would normally review when making decisions? Do you really trust the companies that make AI agents to have your best interests at heart?
An AI agent may have legitimate access to a system and still take an action its user never intended. As these systems become more autonomous, how should responsibility be shared between the developers, the organization deploying the agent, and the people authorizing its actions?
The developer is responsible for building a system with appropriate security protections. The organization deploying it has an obligation to understand all the connections, permissions and risks associated with the agents. And the people authorizing its use need to understand what they’re giving it permission to do.
The fact that an agent made the decision doesn’t change the consequences. At the end of the day, the buck stops with the organizations who have chosen to deploy the AI.
This is why governance needs to be built into the process from the beginning. Clear accountability, strong controls, and visibility into the agent’s actions can prevent a small mistake from turning into a disaster.
One more scary thought: AI is not like a standard program where a developer can easily discover flaws or bugs. AI learns, adapts and grows in ways that the engineers who created it may not anticipate. So, it is getting hard to anticipate issues with AI, or to have the tools to catch those issues.
AI could also help security teams identify exposed personal information and emerging threats much faster. Where do you see the greatest potential for AI in privacy protection, and what safeguards will be needed to ensure these tools do not create new risks through excessive data collection or surveillance?
AI has the greatest potential to help security teams make sense of information quickly. It can make connections between data points, detect patterns indicative of a developing threat, and prioritize cases that need immediate attention. It also has the ability (when done well) to add an abstraction level in transactions. Using an agent to hide who you are, or limit the amount of information you provide, can be very useful.
Additionally, an AI agent can be used to read and understand privacy statements and help you make quick decisions on which sites you can trust more than others. This process is time-consuming today. I would love to see AI agents being used to incorporate privacy into buying decisions. Imagine, for example, that your AI agent recommends product A over product B, not just on price, functionality and reliability, but on privacy. I recently bought a smart TV and soundbar that I wound up having to return when I learned they were collecting my personal information. Agents could help us understand privacy factors that we don’t pay attention to.
Having said that, if you’re using AI to protect people, you need to be sure not to collect or retain more sensitive information than you need. You also need to control who has access to that information, how it’s used, and how long it’s retained. The goal is to reduce people’s exposure and make them safer, not create a detailed profile of their lives.
Ironwall’s work sits at the intersection of digital exposure and real-world safety. When an online risk begins to signal a potential physical threat, what needs to happen between cybersecurity, executive protection, HR, and legal teams to ensure the response is coordinated rather than fragmented?
What a great question. We often tell corporate teams that we are an extension of their executive security and cyber security teams. A good privacy company doesn’t just remove information; it understands when a company is being targeted, when threats are growing from grievances to attacks, and when a pattern of information actually indicates a problem. More importantly, privacy protection gives you room to breathe, so that when a physical or online threat appears, you know that certain information is not exposed and you have more room to deal with the threat.
The problem is that a threat can start in one place and quickly affect several others. For example, a bad actor might publish an executive’s home address, then begin threatening her family, and finally use the information to target her company. If cybersecurity sees one part of this activity, executive protection sees another, and HR or legal sees only the third, the organization can miss the bigger picture. A good privacy protection company helps build channels of communications within a company before threats happen, so that information (not just data) can flow when threats appear.
Companies need a process for sharing relevant threat information across all three teams. If a credible threat emerges online, the physical security team needs to know. If an employee or family member is being targeted, the organization needs to consider their safety as well as potential impact on its systems. Privacy protection resources extend the people on your team and act as a force multiplier, especially when there is a threat. Great companies in this space bring in physical and cyber experts to help your company when emergencies happen.
This means that before an incident even happens, businesses need to put a plan in place that answers questions like: Who assesses a threat? Who makes decisions about physical protection? Who coordinates with law enforcement? Who communicates with the targeted executive?
The goal is to identify the threat early and act before it escalates. Digital exposure and real-world safety can’t be segregated as separate issues when they’re part of the same threat.
Having led Ironwall since 2011, you have seen personal-data exposure evolve alongside the tools used to exploit it. As AI agents become more capable, what assumption about security do you believe business leaders most urgently need to rethink, and what should they put in place today?
Information is necessary, but it is also dangerous. Companies need to understand that information is like electricity — incredibly powerful and useful, but also dangerous. We teach people to understand electricity; we channel it, insulate it and have safeguards in place. We don’t allow children to play near open circuits. Information is the same way. Our personal information can’t be allowed out in the open, available to everyone. It has to be used for the purpose it serves.
AI makes it easier for attackers to create very realistic personalized phishing messages when they have access to information, so we can no longer count on employees to instantly recognize suspicious messages or know fakes from real sites. We need to choke off information so that AI attacks look and feel clumsy enough for us to spot them. AI works on information. Without it, agents are crude and ineffective.
Companies must treat personal data exposure as part of their security risk. This means identifying exactly what information is available, reducing any unnecessary exposure, constantly monitoring for threats, and establishing clear processes for responding when a threat emerges.
Thanks, Ron!
About Ron Zayas
Ron Zayas is the Founder and CEO of Ironwall by Incogni and an expert in online privacy, personal data protection, and security. Since founding Ironwall in 2011, he has focused on helping organizations and at-risk professionals protect sensitive personal information from exploitation, cyber threats, and real-world attacks. A speaker and author, Ron brings extensive experience in privacy protection, executive security, and the evolving challenges of digital safety.
About Ironwall by Incogni
Ironwall by Incogni is an enterprise privacy protection and digital executive security company that helps organizations safeguard executives, employees, and their families from online exposure and targeted threats. The company identifies and removes exposed personal information from across the searchable web while providing continuous monitoring and threat intelligence. By reducing publicly available personal data, Ironwall helps organizations minimize risks associated with identity theft, phishing, harassment, doxxing, and physical and cyber attacks.
