Rockwell Automation Partners with Anthropic on Project Glasswing to Boost AI‑Driven Industrial Cybersecurity

SentiSense · Published Sep 15, 2026, 7:00 AM ET · Updated Sep 28, 2026, 2:45 AM ET

Rockwell Automation said it has joined Project Glasswing, Anthropic's industrial cyber-defense initiative, giving its security teams controlled access to an Anthropic model for finding and remediating vulnerabilities in operational technology. The disclosure came in Rockwell's own Business Wire release on September 15. Glasswing launched in April 2026 with a roughly 50-organization cohort that included AWS, Apple, Google, Microsoft, NVIDIA and Palo Alto Networks, and expanded on June 2 by about 150 more organizations across more than 15 countries. No deal size or deployment timeline was disclosed.

ROK disclosed that it has joined Project Glasswing, Anthropic's collaborative effort to strengthen cyber defense for industrial critical infrastructure using frontier AI models. The announcement came in Rockwell's own Business Wire release, dated Milwaukee, September 15, and quotes Tony Baker, the company's vice president and chief product security officer, saying Rockwell is applying advanced AI capabilities in a controlled, defensive way to identify and address vulnerabilities faster.

One naming detail is worth flagging. Rockwell's release says its security teams gain controlled access to Claude Mythos 5, and the same string appears in foreign-language reproductions of the release. Anthropic's own Glasswing pages do not use a numbered model name, referring throughout to Claude Mythos Preview. The model name in this story is therefore Rockwell's, as published, rather than Anthropic's branding.

The program context matters for reading the announcement's weight. Glasswing launched in April 2026 with an initial cohort of roughly 50 organizations, named on Anthropic's site as Amazon Web Services, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks, alongside program-wide commitments of up to $100 million in usage credits and $4 million in donations to open-source security organizations. On June 2 Anthropic said the program had added approximately 150 more organizations across more than 15 countries without naming them, which is consistent with Rockwell's membership surfacing through Rockwell rather than through an Anthropic list.

For Rockwell the practical question is whether this moves beyond a security-research collaboration into product. No dollar figure specific to Rockwell was disclosed, no deployment timeline was given, and the language in the release is exploratory. Operational technology networks are a genuinely underserved target for automated vulnerability discovery, so a credible result here could differentiate Rockwell's security portfolio against automation peers. What to watch is whether Rockwell publishes measured outcomes, such as incident response times or disclosed fixes, and whether any of it reaches the products its industrial customers buy.