Daniel Schwartz, President of Design I.T. Solutions, LLC, a leading consulting, cybersecurity and managed IT services provider.getty​The AI race is on, and vendors are accelerating their push to the top. I believe we have a right to be concerned about the impact of AI on our well-being. However, it's not a droid army coming for us. It's much simpler than that.To understand what's happening, we need to look back at the Industrial Revolution. In many ways, today's AI-driven automation is a modern version of it. That era brought massive growth through automation and new ways of performing old tasks. Manufacturing improved, and products reached the market faster than ever before.That progress came with a cost, though. Early factories often lacked basic safety measures. Conveyor belts were installed without guardrails. Because quality controls and workplace protections hadn't caught up with the pace of innovation, workers were placed in physically dangerous environments.I believe we're seeing a similar pattern emerge with AI. As vendors race to outpace competitors with faster, more capable releases, safeguards can lag behind. Some AI systems have been accused of failing to appropriately respond to users in crisis—potentially pointing to missing safeguards.We're also seeing concerns between AI-assisted automation and vehicle-related incidents that have resulted in perilous situations. In healthcare, using AI to diagnose conditions or guide self-treatment can lead to misdiagnosis, delayed care or dangerous outcomes. None of this is an argument against AI adoption. The benefits are too significant to ignore. Rather, it's a clear signal that we need to pay close attention to how these systems are being built, tested and deployed.From a cybersecurity perspective, the stakes are even higher. As AI becomes embedded in healthcare—powering large-scale data models for disease research, diagnosis and treatment planning—we must ensure the integrity of that data. A compromised system can create widespread harm.The same applies to infrastructure. As AI begins to control elements of power grids, transportation and other critical systems, the risk shifts from theoretical to tangible. If attackers gain access and manipulate AI-driven controls, the consequences could directly impact cities and public safety.This all points back to one thing: quality control.We've seen what happens when it's overlooked. The supply chain attack involving SolarWinds exposed how a vulnerability could impact thousands of organizations. Now imagine that embedded in widely adopted AI systems. The scale—and the potential fallout—would be far greater.We can't stop the march of AI integration, and we shouldn't try to. What do we do about it?​As business leaders, we have to be more deliberate in how we adopt AI. That starts with vendor scrutiny. We should be asking hard questions: Which quality controls are in place? How are human safety scenarios handled? Which cybersecurity measures protect the code, models and data? The answers shouldn't be pie in the sky, vague or overly complicated.On quality control, I want to see vendors clearly explain their testing procedures. That includes attack model testing (intentionally trying to break the model) and validation processes before and after each release. If they can't show and explain how they test for edge cases and failure scenarios, that's a red flag.On human safety, there should be defined guardrails for defined high-risk interactions. This includes escalation paths when an interaction presents signs of harm, medical distress or potential sensitive situations. AI shouldn't be the final authority in those moments; it should be designed to defer, redirect or involve human support.On cybersecurity, the baseline needs to go beyond standard application security. We should be asking how models are protected from tampering, how training data is validated and how updates are secured. Controls around access, code integrity and supply chain security should be clearly documented and independently tested.These expectations aren't new. Established frameworks such as SOC 2, CMMC and ISO/IEC 27001 already require this level of discipline. If a vendor can't clearly explain how they prevent unauthorized model manipulation or poisoned data inputs, you're taking on unnecessary risk.Finally, accountability matters. Vendors should have a published and clearly defined process for incident response, including how they communicate issues, remediate them and prevent recurrence. Transparency in this area is just as important as the technology itself.Vendor scrutiny is only part of the equation. Organizations also need to build their own internal guardrails to ensure AI is used safely and responsibly across the business.Like everything in IT, it starts with employee training—but not in the traditional "check the box" sense. Teams need practical, role-based guidance on how AI should and shouldn't be used in their day-to-day work. That includes understanding where AI can help productivity but also where it can introduce risk—such as handling sensitive data, making assumptions based on incomplete outputs or over-relying on AI-generated recommendations without validation.From a policy standpoint, organizations should establish clear usage guidelines early. Define what data can be entered into AI systems, what cannot and under what circumstances. Sensitive information—whether it's customer data, healthcare records, financial data or intellectual property—should never be exposed to AI tools without proper controls in place. If that line isn't clearly drawn, it will be crossed.There also needs to be a layer of human accountability. AI should support decision making, not replace it. Organizations should implement procedures that require human review for high-impact outputs—whether that's customer communications, financial decisions or operational changes driven by AI insights.​AI is a powerful tool, a resource and, increasingly, a business partner. It should be embraced, but that adoption must come with expectations. Vendors need to prioritize safety and security alongside innovation—not after it.The lessons from the Industrial Revolution are still relevant. Progress without guardrails can lead to preventable harm. If we keep those lessons in mind, we can avoid repeating the same mistakes as we build this next era of technological advancement.​Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?