Reflection joins Texas water-security pilot to help build locally run AI defenses
The company says locally deployed models can keep sensitive operational data out of third-party services. The pilot also targets utilities short on cybersecurity staff and funding.
Listen to this story
The audio brief
Story brief
3 key pointsReflection is contributing to Project Watershed 250, a Texas pilot launched in August, to develop locally deployable AI cyber defenses for water and wastewater utilities. Its open-weight approach is intended to let operators inspect and adapt models, keep operational data off third-party services, and reduce reliance on providers that could cut access; no security gains have yet been demonstrated. Texas Cyber Command says participating firms will provide expertise and tools at no cost, targeting a sector with many utilities that lack resources for specialized cybersecurity support.
- 01
Texas Cyber Command named Reflection among the pilot’s first 63 industry partners on November 17, three days before Reflection announced its participation.
- 02
Texas has more than 7,400 public water systems and 3,000 wastewater treatment facilities, according to Texas Cyber Command.
- 03
Program-wide offerings may include security assessments, technical assistance, and remediation tools; utilities can choose services to match their needs.
Reflection is joining a Texas effort to protect water and wastewater systems with AI tools that utilities could run locally. The company’s November 20, 2026, announcement commits it to helping develop cyber defenses through Project Watershed 250, with a focus on giving operators control over the technology and keeping sensitive operational data off third-party services.
Texas Cyber Command had already named Reflection among the pilot’s first 63 industry partners on November 17. The initiative launched in August through the White House Office of the National Cyber Director, working with the Texas governor’s office and Texas Cyber Command.
Control over data—and continued access
Reflection’s pitch centers on open-weight models: AI systems that utilities and their vendors can inspect, adapt and deploy locally. The company says this approach lets operators shape the technology around their own needs. It also gives them room to evaluate a tool’s capabilities and limitations before deciding where and how to use it.
Beyond data privacy, the company points to continuity of access. Operators may worry that a third-party provider could abruptly withdraw access to a system they use. Reflection presents locally deployable models as an answer to that dependence.
Free support for utilities with fewer resources
Texas Cyber Command says the state has more than 7,400 public water systems and 3,000 wastewater treatment facilities. The pilot is designed for utilities of different sizes, including providers that lack the staff, funding or technical expertise to obtain specialized cybersecurity support themselves.
Participating companies are volunteering expertise, tools and personnel at no cost to utilities. Texas Cyber Command says the services may include:
- Cybersecurity assessments and technical assistance.
- Tools and remediation support to address security problems.
Those are program-wide offerings, rather than a description of Reflection’s contribution alone. Utilities can select services according to their own needs and priorities. The first group of 63 companies is not a closed roster: Texas Cyber Command says the pilot remains open to additional industry applicants.
Both Reflection and Texas Cyber Command see potential beyond this pilot. Reflection hopes the work will help shape AI-based infrastructure defense nationally. The agency describes a framework that, if successful, could extend to other infrastructure sectors in Texas and other states—not an announced expansion.
Sources
- txcc.texas.gov63 Companies Step Up to Defend Texas Water Systems in Project Watershed 250 | Texas Cyber Command
- reflection.aiReflection joins initiative to defend Texas water systems with open-weight AI models — Reflection
Reader comments
Newest comments first. Replies stay oldest first.