Rockwell Automation has joined Anthropic’s Project Glasswing, bringing frontier artificial intelligence into the fight to secure industrial software, connected manufacturing systems and critical infrastructure.
The announcement is another important sign that AI is moving beyond productivity tools and code generation into one of the most demanding areas of industrial technology: cybersecurity and operational resilience.
Rockwell Automation announced on 15 September 2026 that it is participating in Anthropic’s Project Glasswing, an initiative designed to help organisations responsible for critical software identify and address security vulnerabilities using advanced AI models.
What is Project Glasswing?
Anthropic launched Project Glasswing in April 2026 after observing a significant increase in the cybersecurity capabilities of its frontier AI models.
The initiative gives approved organisations controlled access to advanced Claude models for defensive cybersecurity work. Initial participants included major technology and infrastructure organisations such as Microsoft, Cisco, Google, NVIDIA, AWS, CrowdStrike and Palo Alto Networks.
The objective is straightforward but significant: use increasingly capable AI systems to find vulnerabilities before attackers do.
Anthropic reported in May that Project Glasswing and its initial group of approximately 50 partners had already identified more than 10,000 high- or critical-severity vulnerabilities. The programme was subsequently expanded to approximately 150 additional organisations across more than 15 countries.
How Rockwell Automation Plans to Use AI
According to Rockwell Automation, its security teams are exploring the use of Claude Mythos 5 to accelerate several stages of vulnerability management, including:
- vulnerability discovery;
- validation of potential security issues;
- prioritisation according to risk;
- remediation of identified weaknesses; and
- security analysis across connected software and industrial systems.
The work is being carried out through controlled access and is focused specifically on defensive cybersecurity.
For manufacturers, this is particularly relevant.
A modern factory is no longer an isolated collection of PLCs, HMIs and drives. Manufacturing environments increasingly connect production equipment with SCADA platforms, historians, MES systems, cloud services, remote-access infrastructure and enterprise IT.
That connectivity creates enormous opportunities for optimisation and automation—but it also increases the number of systems that must be protected.
Why This Matters for Industrial Automation
Traditional vulnerability analysis is heavily dependent on specialist knowledge and human engineering time.
Advanced AI introduces the possibility of analysing very large codebases, binaries and interconnected software systems at a scale that would be extremely difficult to achieve manually.
Anthropic says its Mythos-class models have demonstrated the ability to identify previously unknown vulnerabilities in major operating systems, browsers and other important software, including weaknesses that had remained undiscovered for many years.
In an industrial environment, the potential implications are substantial.
AI-assisted cyber defence could eventually help engineering and security teams analyse firmware, industrial applications, communication services, engineering tools and supporting infrastructure more quickly—allowing vulnerabilities to be discovered and corrected earlier in the lifecycle.
The important shift is from reacting to known vulnerabilities towards continuously searching for weaknesses before they become exploitable incidents.
IT Security and OT Security Are Converging
Industrial cybersecurity has historically been treated differently from conventional IT security.
And for good reason.
A failed office application may interrupt business operations. A compromised industrial control system could stop production, damage equipment or affect essential services.
Availability, deterministic behaviour, safety and long equipment lifecycles all make operational technology environments different from normal enterprise IT.
However, the boundary between the two environments continues to disappear.
Industrial Ethernet, OPC UA, edge computing, remote engineering, virtualisation and cloud-connected production systems increasingly connect the OT and IT domains.
That means security strategies also need to evolve.
Rockwell’s participation in Project Glasswing suggests that major industrial automation vendors increasingly see advanced AI as part of that next generation of defence.
AI Creates Both Opportunity and Risk
There is also an uncomfortable reality behind initiatives such as Project Glasswing.
The same AI capabilities that help security researchers discover vulnerabilities can potentially be used by attackers.
Anthropic created Project Glasswing partly because its research showed that frontier models were becoming significantly more capable at vulnerability discovery and exploit development. The organisation’s strategy is therefore to provide capable defenders with early access to these technologies while developing additional safeguards around their use.
This could create a new cybersecurity dynamic.
Instead of security teams manually searching for weaknesses while attackers automate more of their activity, both sides may increasingly use AI.
The advantage may belong to organisations that integrate AI-assisted security into their engineering processes early.
What Should Manufacturers Take From This?
Manufacturers do not need access to frontier cybersecurity models tomorrow to respond to this trend.
But several principles are already becoming increasingly important:
- build industrial systems using secure-by-design principles;
- maintain accurate inventories of connected OT assets;
- segment industrial and enterprise networks correctly;
- control and audit remote engineering access;
- maintain firmware and software vulnerability-management processes;
- treat industrial software development with the same security discipline as modern IT development;
- monitor connections between PLC, SCADA, MES, cloud and enterprise systems; and
- prepare for AI to become part of both cyber defence and cyber threats.
Cybersecurity can no longer be treated as an additional layer added after an automation project has been commissioned.
It increasingly needs to be considered alongside control architecture, functional safety, networking, software engineering and lifecycle support.
The Bigger Picture
Rockwell Automation joining Project Glasswing is significant because it connects two technology trends that are rapidly reshaping manufacturing: industrial digitalisation and frontier AI.
We are likely to see AI used increasingly across the entire automation lifecycle—from requirements analysis and PLC software generation to commissioning, diagnostics, predictive maintenance and cybersecurity.
For automation engineers, system integrators and manufacturers, the question is therefore becoming less about whether AI will enter industrial engineering and more about where it can be introduced safely, responsibly and with measurable value.
At Antomatix, we believe the most valuable industrial AI systems will augment experienced engineers rather than attempt to remove engineering expertise from the process.
Cybersecurity is another area where that combination of human engineering judgement and AI-scale analysis may prove particularly powerful.
Source: Rockwell Automation, Rockwell Automation Joins Anthropic’s Project Glasswing to Advance AI-Enabled Cyber Defense for Industrial Critical Infrastructure, 15 September 2026; Anthropic Project Glasswing announcements.