AI Governance in 2026: What Enterprise Teams Must Solve

As we come to the close of 2025, one of the clearest observations is just how quickly the data and AI landscape is changing. The push for AI in production is stronger than ever. Data environments continue to grow in size and complexity. AI products are steadily moving from experimentation into maturity. Along the way, expectations around trust, reliability, and governance have continued to rise.

Throughout 2025, Foundational continued to strengthen data and AI governance for the AI era by focusing our efforts where change actually happens, at the source code. And there’s a lot more code now. This approach is core to how Foundational delivers proactive governance across enterprise environments (learn more). The focus was not on chasing trends, but on reinforcing what durable governance looks like when data and AI are central to how organizations run.

This post reflects on the milestones we delivered in 2025, the patterns we observed across customer environments, and how those signals are shaping what data and AI governance needs to support in 2026.

‍

What We Focused on Building in 2025

Governance that fits real enterprise environments

A consistent priority throughout the year was ensuring governance works across the environments teams actually operate, from established on-premises databases to the newest cloud systems. Enterprise tech stacks span multiple platforms, services, databases, analytics tools, and increasingly, AI systems. Governance needs to function across all of them without adding friction.

Key milestones in this area included:

General availability support for GitLab, Bitbucket, and Azure DevOps, making us compatible with every source code platform a large company might be using today. Git integrations.
Redesigned lineage computation to enterprise scale, supporting accurate impact analysis across complex data systems for hundreds of thousands of objects and more, while keeping clarity and ease of use.
Improved issue detection and root cause analysis to help teams move from reactive to proactive, faster.

These investments reflect a simple principle. Governance is most effective when it aligns with existing workflows rather than requiring new ones.

‍

Making data quality continuous and predictable

Another focus area was data quality. Many teams rely on point in time checks or alerts tied to specific warehouse tables. What we continued to see was the value of more precise signals that reflect how data is expected to behave over time, and the ability to scale that to large environments while minimizing alert fatigue

Key improvements included, building on Foundational’s data quality and observability capabilities:

Fully automated anomaly checks across warehouses and databases.
Early, pre-merge visibility into changes, before downstream symptoms appear.

Pre-deploy semantic checks extended that visibility earlier in the development lifecycle. This made it easier to understand the impact of change before it reached production, reducing rework and shortening feedback loops.

‍

Context as the foundation for trust

As data environments grow, context becomes increasingly important. Monitoring alone cannot answer the questions teams care about most. What changed. Why it matters. Who is affected.

Throughout the year, we expanded the Foundational Data Graph with a focus on:

Richer understanding of how systems are connected.
Anomaly detection improvements that reduced noise.
Broader lineage coverage across SQL and non SQL systems.

This shared context supports clearer conversations and better decisions across teams. You can explore how this works in Foundational’s lineage analysis experience, which shows how source code level lineage drives proactive governance across the data stack.

‍

Governance closer to where work happens

We also continued to invest in governance that integrates directly into the tools and workflows teams already use.

Code based lineage expanded into additional languages such as Java and C++, new orchestration frameworks, Python notebooks, and non relational databases such as MongoDB and Couchbase. This made it easier to understand change across the full lifecycle, from source code through downstream consumption.

At the same time, we’ve made lineage, metadata, and documentation available directly within existing BI tools. This helped business users build confidence in the data they rely on without needing to leave their existing workflows.

‍

Extending governance foundations to AI

AI adoption continued to accelerate in 2025, though teams remain at different stages of maturity. Rather than treating AI governance as a separate discipline, our focus has been on extending existing governance foundations to support AI systems, including production AI workflows that depend on trusted data and governed change:

If before, the data organization needed lineage to have a single source of truth for data flows, it is now the CIO and CTO who need visibility into every AI flow in the company, particularly the ones that may access sensitive data.
If before, CI/CD was useful for flagging data issues and breaking schema changes, it is now even more crucial for having guardrails to prevent issues when 30% or more of the code is written or co-written by AI.

All the core data governance fundamentals, including lineage, data quality, access controls, and policy enforcement, play a role in AI governance as data flows into models and applications. Early AI assisted workflows helped teams scale tasks like documentation and discovery while keeping humans in the loop.

‍

Signals That Shaped the Road Ahead

Industry research throughout the year reinforced several consistent themes, echoed by leading analyst firms such as Forrester, Gartner, and 451 Research:

Trust and explainability are increasingly central to AI outcomes.
Governance plays a critical role in turning AI investment into durable value.
As AI systems become more embedded in core operations, the need for visibility and control continues to increase.

Together, these signals point to a growing consensus around the importance of foundational governance capabilities, even as organizations take different paths to maturity. For example, Forrester’s Predictions 2026 highlight trust and governance as decisive factors for scaling AI responsibly (Forrester Predictions 2026).

‍

Patterns We See Across Teams Making Progress

While approaches vary, several patterns appear consistently among teams navigating complexity with confidence.

They tend to:

Use lineage as a shared reference point for understanding systems.
Rely on metadata and schema intelligence rather than manual documentation.
Introduce governance earlier in the lifecycle, closer to where changes are made.
Use automation thoughtfully to reduce repetitive work while maintaining oversight.

Looking Ahead to 2026: Is Lineage Hot Again?

In 2026, governance will continue to mature alongside data and AI systems. We expect to see deeper lineage across data and AI workflows, richer context for understanding change, and governance capabilities that extend further into development, analytics, and AI tooling. The need for lineage that spans across tools, applications, and systems to provide full visibility is stronger than ever. And in that universe, Foundational’s code analysis is key.

As AI becomes more integrated into everyday operations, governance will increasingly serve as an enabling layer that helps teams move forward with confidence.

‍

The Biggest AI Governance Challenges for Enterprise Teams in 2026

Three challenges show up consistently across the enterprise teams we work with heading into 2026.

Provenance for model inputs. Model risk teams can explain what a model does. Fewer can explain, with evidence, exactly which data fed a given model version and where that data originated. SR 26-2 and OCC Bulletin 2026-13 both raise the bar on this specific gap for regulated institutions, and the same evidence expectation is spreading to less regulated industries as boards ask the same question.

AI generated code entering the pipeline. A growing share of pipeline code is now written or co-written by AI. That code changes schemas and joins as confidently as human written code, and often faster, which means governance checks that used to run on a human review cadence now need to run on every commit.

Governance that stops at the warehouse. Catalog tools see only where data lands. They do not see the Python pipelines, ORM layers, and application code that produced it. When a model input traces back through an application layer a catalog never indexed, the lineage trail ends exactly where the audit question begins.

Foundational is a data and AI governance platform built to close that gap. By analyzing source code directly rather than warehouse query logs, deterministic lineage extends upstream through the pipelines and application code that catalog based tools cannot see.

Closing Thoughts

2025 reinforced an important lesson. Durable governance is built intentionally, grounded in real systems, and designed to scale with change.

At Foundational, we are building a data and AI governance platform with governance anchored at the source code rather than applied after the fact, using source code analysis to see what catalog tools cannot. That approach is what trusted, scalable data and AI operations will require in 2026 and beyond.

Frequently Asked Questions

What are the biggest AI governance challenges for enterprise teams in 2026?

The three that come up most are proving provenance for the data feeding AI models, keeping pace with AI generated changes to data pipelines, and closing the visibility gap left by catalog tools that only see data once it reaches the warehouse. Regulatory guidance like SR 26-2 and OCC Bulletin 2026-13 has made the first challenge a compliance requirement rather than a best practice for regulated industries.

How is AI governance different in 2026 than it was in 2025?

In 2025, most organizations were still establishing basic lineage and metadata practices. In 2026, those practices are being tested directly against AI specific requirements: model input provenance, agent reliability, and evidence that governance holds up when a meaningful share of pipeline code is AI generated rather than human written.

What does SR 26-2 require for AI model governance?

SR 26-2 superseded SR 11-7 in April 2026 and narrowed the formal scope of model risk management for banks, explicitly excluding generative and agentic AI from that formal scope. That does not remove the governance obligation. It shifts responsibility for governing generative and agentic AI models onto the institution's own data and AI governance program rather than the model risk framework.

Why do catalog based tools fall short for AI governance?

Catalog tools index data after it lands in the warehouse, which means they can only show where data ended up, not where it originated. AI governance requires tracing a model input back through the Python pipelines, ORM layers, and application code that produced it, and that is source code catalog tools never analyze.

What should an enterprise AI governance program prioritize first in 2026?

Start with provenance for the data feeding the highest risk models, the ones used in regulated decisions or customer facing agents. Establish deterministic lineage back through the full pipeline and application layer for those models first, then extend coverage outward rather than trying to govern every dataset at once.

code snippet <goes here>

<style>.horizontal-trigger {height: calc(100% - 100vh);}</style>
<script src="https://cdnjs.cloudflare.com/ajax/libs/gsap/3.8.0/gsap.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/gsap/3.8.0/ScrollTrigger.min.js"></script>
<script>
// © Code by T.RICKS, https://www.timothyricks.com/
// Copyright 2021, T.RICKS, All rights reserved.
// You have the license to use this code in your projects but not to redistribute it to others
gsap.registerPlugin(ScrollTrigger);
let horizontalItem = $(".horizontal-item");
let horizontalSection = $(".horizontal-section");
let moveDistance;
function calculateScroll() {
  // Desktop
  let itemsInView = 3;
  let scrollSpeed = 1.2;  if (window.matchMedia("(max-width: 479px)").matches) {
    // Mobile Portrait
    itemsInView = 1;
    scrollSpeed = 1.2;
  } else if (window.matchMedia("(max-width: 767px)").matches) {
    // Mobile Landscape
    itemsInView = 1;
    scrollSpeed = 1.2;
  } else if (window.matchMedia("(max-width: 991px)").matches) {
    // Tablet
    itemsInView = 2;
    scrollSpeed = 1.2;
  }
  let moveAmount = horizontalItem.length - itemsInView;
  let minHeight =
    scrollSpeed * horizontalItem.outerWidth() * horizontalItem.length;
  if (moveAmount <= 0) {
    moveAmount = 0;
    minHeight = 0;
    // horizontalSection.css('height', '100vh');
  } else {
    horizontalSection.css("height", "200vh");
  }
  moveDistance = horizontalItem.outerWidth() * moveAmount;
  horizontalSection.css("min-height", minHeight + "px");
}
calculateScroll();
window.onresize = function () {
  calculateScroll();
};let tl = gsap.timeline({
  scrollTrigger: {
    trigger: ".horizontal-trigger",
    // trigger element - viewport
    start: "top top",
    end: "bottom top",
    invalidateOnRefresh: true,
    scrub: 1
  }
});
tl.to(".horizontal-section .list", {
  x: () => -moveDistance,
  duration: 1
});
</script>

2025: What We Learned Building Data Governance for the AI Era