The cybersecurity conversation shifted the moment Anthropic introduced Claude Mythos under Project Glasswing in April 2026. Positioned as a breakthrough in AI-driven security, Mythos demonstrated capabilities once reserved for elite researchers—discovering zero-day vulnerabilities, chaining exploits across systems and generating functional proof-of-concept code with minimal guidance.
For security leaders, the implications were immediate. Gartner's 2026 information security forecast names "securing AI" the fastest-accelerating investment category through 2030, while its top trends report places cybersecurity oversight of Agentic AI at the center of the modern CISO agenda. Enterprise IT has largely adopted the shift. Operational technology (OT) environments, however, have responded with far greater caution and for good reasons.
Why OT teams remain cautious
Unlike enterprise IT, OT environments run critical infrastructure where availability, reliability and safety are non-negotiable. Refineries, power substations, manufacturing plants, water utilities and pharmaceutical production lines cannot absorb experimentation that introduces operational risk. Several concerns continue to shape how these teams view frontier AI.
The threat landscape is already expanding. Industrial sectors are seeing a rise in cyber incidents, and critical infrastructure operators face mounting pressure from sophisticated adversaries, so adding autonomous AI to the mix naturally raises difficult questions about risk. Those questions sharpen as nation-state actors begin folding advanced AI into offensive operations. As AI lowers the technical barrier to sophisticated attacks, the comforting assumption that industrial systems can only be targeted by rare specialists becomes almost impossible to maintain.
There is also a cultural dimension. OT organizations are built on safety-first principles and the engineers responsible for industrial control systems are understandably reluctant to hand autonomous agents decision-making authority over live environments. Human oversight is not optional. Governance is still catching up, too—traditional identity, access management and security controls were designed for human users, not autonomous agents operating at machine speed. Viewed this way, OT skepticism is a rational response rooted in operational reality.
The opportunity hidden inside the skepticism
The organizations that pull ahead will not treat frontier AI as something to avoid. They will use it as a force multiplier for defense while applying the controls needed to preserve operational trust. Frontier AI can help defenders find vulnerabilities in legacy systems before attackers do and it can analyze attack paths across Purdue Model layers to reveal how seemingly isolated weaknesses could be chained into serious business risk. It also transforms prioritization: Rather than drowning teams in thousands of findings, it surfaces the small subset that poses the greatest operational threat. As enterprises introduce digital twins, remote-access solutions, advanced analytics and AI-enabled workflows, the same capability supports continuous validation, which involves constantly assessing whether new risks have emerged. Most importantly, AI helps defenders maintain parity. If adversaries gain advanced AI, defenders need comparable tools to identify, analyze and respond at the same pace.
Converting OT concerns into security advantages
OT skepticism is so valuable precisely because every concern, addressed deliberately, converts into a strategic advantage:
- Slow patch cycles can drive wider adoption of virtual patching and compensating controls
- Concerns about increasingly capable attackers can accelerate investment in AI-driven detection and response
- Worries about model exposure can lead to stronger governance and sovereign deployment strategies
- A safety-first culture becomes the foundation for human-in-the-loop AI architectures
- Governance gaps create the impetus for agent-aware identity, inventory and access controls
The result is not less innovation, but a controlled approach to it.
HCLTech 360° SecureOT for AI-powered enterprises
This is exactly what HCLTech 360° SecureOT was built to enable—a dynamic, programmatic OT/IIoT cybersecurity framework aligned to ISA/IEC 62443, NIST SP 800-82, SANS CSC and ISA-95, designed to help clients safely convert frontier AI capability into OT resilience.
Built on three pillars of securing the Purdue Model, Zero Trust for OT and protection of critical infrastructure, we support clients adopting Mythos-class AI in and around their OT estate with AI-augmented asset discovery and an OT security operations center featuring passive discovery and ICS-tuned threat detection. We apply frontier-AI-informed threat modeling to preemptively identify chained exploit paths across the Purdue layers, enforce safe-by-design AI guardrails with strict human-in-the-loop control and close the vector attackers reach for first through Zero Trust and secure remote access. And because containment and safe-state recovery matter as much as detection in OT, we build resilience and recovery playbooks into the model from the start.
Our broader AI-powered cybersecurity portfolio extends this further, securing AI models, prompts, APIs and enterprise data while supporting governance, compliance, explainability and autonomous security operations. It is an integrated approach that reflects the growing need for unified protection across converged IT and OT environments.
The bottom line
Mythos is neither an existential threat nor a universal cure. It is a new class of cybersecurity capability that will increasingly shape both offensive and defensive operations. For OT leaders, the question is how to adopt it responsibly. The organizations that succeed will replace fear with governance, hype with engineering discipline and uncertainty with trust built on human oversight, operational controls and proven security frameworks. Seen in that light, OT skepticism is not an obstacle to AI adoption. It is the foundation for doing it safely.
