Artificial intelligence is moving closer to the physical layer of industrial automation, and new research published in August 2026 is examining what could happen when autonomous AI agents interact with real programmable logic controllers.

The research introduces a new hardware-in-the-loop framework called PLCBENCH. Instead of evaluating AI only through software simulations, the framework connects AI agents with commercial PLC hardware and closed-loop industrial process simulations.

Researchers evaluated five families of large language models across 240 real-PLC test episodes involving four commercial PLCs and four closed-loop workloads.

The results demonstrate an important development for the automation industry. In a portion of the experiments, AI agents were able to move beyond basic PLC interaction and achieve sustained physical objectives within the simulated industrial processes.

This does not mean that AI systems are ready to independently operate factories. However, the research shows why the relationship between AI and industrial control systems deserves closer attention.

A PLC is fundamentally different from an ordinary computer. It can directly control motors, pumps, valves, conveyors, heaters and other physical equipment. If an unauthorized system gains access to PLC logic or process parameters, the consequences can potentially extend from digital systems into the physical environment.

At the same time, AI could provide significant benefits for legitimate industrial applications.

AI assistants could help automation engineers analyze PLC diagnostic information, review alarm conditions, generate documentation, identify potential programming errors and support troubleshooting. AI could also assist with legacy system modernization by helping engineers understand older PLC programs and relationships between controllers, I/O modules, HMIs and communication networks.

AI-assisted PLC programming is another emerging area. Engineers may increasingly use AI tools to generate portions of control logic, but automatically generated code still requires testing, validation and commissioning before being deployed to production equipment.

This is particularly important in safety-critical industrial environments. A program that appears logically correct may behave differently when connected to real sensors, actuators and process conditions.

The new PLC research therefore highlights two sides of industrial AI. On one side, autonomous agents could create new cybersecurity challenges when they gain access to operational technology systems. On the other side, controlled AI applications could help engineers improve productivity, maintenance and automation development.

For PLC and DCS professionals, the key issue will be finding the right balance between automation intelligence and engineering control.

As AI becomes more capable of interacting with industrial devices, future automation architectures may combine traditional PLC and DCS technologies with AI-based analysis and decision-support systems.

The development of AI-powered industrial automation is still at an early stage, but recent research suggests that the boundary between artificial intelligence and physical automation is becoming increasingly important.

For the global PLC, DCS and industrial automation industry, AI is likely to become one of the major technologies shaping the next generation of control systems.