Operational-technology providers express annoyance over Mythos exclusion

Nextgov reports that operational technology (OT) providers and industry groups have pressed Anthropic for access to the Mythos Preview cybersecurity model after an initial rollout focused on major tech and finance firms, according to four people familiar with the discussions. Nextgov says OT representatives voiced frustration during recent roundtables and listening sessions about being left out of Project Glasswing, Anthropic's initiative with major companies designed to secure critical software. Two of the sources told Nextgov the processes for granting OT firms access are ongoing. Nextgov reports that one participant said, "There's definitely an annoyance in the OT world." Nextgov has asked Anthropic, the Office of the National Cyber Director, and American Water for comment. The article frames the disagreement as highlighting challenges to building an industrywide push to patch cyber vulnerabilities affecting critical infrastructure.
What happened
Nextgov reports operational technology (OT) providers and their trade groups have pushed for access to Anthropic's cybersecurity-focused Mythos Preview model after the model's initial rollout prioritized major technology and finance firms, according to four people familiar with the discussions. Nextgov says OT representatives have expressed frustration in recent roundtables and listening sessions about being left out of Project Glasswing, Anthropic's initiative with major companies to secure critical software globally. Two sources told Nextgov that processes to grant OT firms access are ongoing. Nextgov reports that one attendee said, "There's definitely an annoyance in the OT world." Nextgov has asked Anthropic, the Office of the National Cyber Director, and American Water for comment.
Technical details
Per Nextgov reporting, Mythos Preview is being used within Project Glasswing as part of a coordinated vulnerability-patching effort among selected large firms. The article does not provide further technical specifications of Mythos Preview or the integration mechanisms being applied to OT systems.
Industry context
Editorial analysis - technical context: Companies and sectors that operate legacy OT stacks typically face high variation in protocol support, patching windows, and asset visibility. Industry-pattern observations note that aligning AI-driven vulnerability detection and automated patch workflows with OT environments usually requires tailored connectors, extended testing windows, and close coordination with vendors and operators.
Context and significance
Editorial analysis: The Nextgov report highlights an access and governance tension that can arise when new security tooling is distributed first to a subset of large enterprises. Observed patterns in similar initiatives show that excluding OT vendors or operators from early testing can delay operational validation on industrial control systems and slow adoption of mitigations across critical-infrastructure supply chains.
What to watch
For practitioners:
- •Which OT vendors or industry associations gain formal access to Mythos Preview and under what data-sharing terms.
- •Whether the Office of the National Cyber Director or other coordinating bodies issue guidance tying AI-assisted patching pilots to OT-specific safety checks.
- •How integration testing addresses OT constraints such as maintenance windows, safety-critical fail-safes, and vendor firmware lifecycle.
Nextgov is the source for the reported facts in this piece; where the reporting cited anonymous participants, their observations are attributed to those sources.
Scoring Rationale
This story matters to security and infrastructure practitioners because access barriers to an AI-driven vulnerability tool could delay validation and deployment for OT environments. It is notable but not paradigm-shifting: the issue centers on governance and integration rather than a new technical capability.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
