The AI Accountability Report from GitLab, based on a survey of 1,528 DevSecOps professionals, examines how organizations are adopting AI in software development and the emerging challenges around accountability, governance, and maintainability.
What is AI Accountability for those who don’t know? AI Accountability refers to the organizational and technical ability to answer three fundamental questions about AI-generated code and agentic workflows: Where did it come from? What was it meant to do? And who is responsible for it once it’s in production?
The report shows that AI coding tools have moved from experimentation to core infrastructure. Most organizations now use multiple AI tools in production, with strong reported gains in developer productivity and code quality. However, this acceleration has shifted bottlenecks downstream into code review, security scanning, compliance, and deployment phases. Long-term maintainability of AI-generated code has become a top concern, with 73% of respondents worried about technical debt accumulation.
Context and traceability are highlighted as the new competitive advantages. As code generation becomes more commoditized, the ability to maintain connected context across the full software lifecycle and trace code back to its original intent and business requirements will separate leading organizations from the rest. Governance is also rising as a priority, with many teams recognizing that adoption has outpaced formal policies and controls.
The report emphasizes that sustainable AI success in software development will depend less on how fast code can be generated and more on how effectively it can be governed, traced, and maintained at scale.
Key Findings
AI coding tools have become mainstream infrastructure, with 91% of organizations using two or more AI coding tools in active production and 54% using three or more.
ROI from AI coding tools is exceeding expectations, as 60% of DevSecOps professionals report that their organization’s investment has outperformed expectations, with only 9% saying it has fallen short.
Developer productivity has improved significantly, with 78% of respondents saying individual developers write and commit code faster since adopting AI coding tools.
Overall code quality reaching production has also improved for 73% of organizations since adopting AI coding tools.
Bottlenecks have shifted downstream, as 85% agree that AI has moved the bottleneck from writing code to reviewing and validating it.
Long-term maintainability of AI-generated code is a major concern, with 73% of DevSecOps professionals expressing worry and 82% agreeing it risks creating a new form of technical debt.
Context and traceability are emerging as critical differentiators, with many organizations struggling to connect AI-generated code back to original business requirements and track its origin.
Governance adoption is growing but still maturing, with three-quarters of organizations having started building governance frameworks for AI-generated code.
Formal governance policies lead to better outcomes, as organizations with formal policies report higher rates of human review for AI-generated code compared to those without.
Tool integration across the software lifecycle remains limited, with only 28% of organizations saying their tools are fully integrated with shared data and workflows.
The majority of organizations (84%) agree that the biggest challenge with AI-generated code is governing what happens to it after it is created.
DevSecOps professionals overwhelmingly agree (85%) that the next phase of AI in software development will focus less on generating code and more on governing and maintaining it.
What the Report Covers
The AI Accountability Report from GitLab, conducted with The Harris Poll, is based on a survey of 1,528 DevSecOps professionals from six countries across North America, Europe, and Asia-Pacific. The report provides a comprehensive look at how organizations are adopting AI coding tools, the resulting shifts in the software development lifecycle, and the emerging challenges around context, traceability, governance, and long-term maintainability of AI-generated code.
The report is organized into four main parts plus supporting sections:
AI is Working — and Reshaping the Software Lifecycle — This section details mainstream adoption of AI coding tools (91% of organizations use two or more), strong ROI performance (60% say it exceeded expectations), productivity gains (78% report faster code writing and committing), and improved code quality (73%). It also highlights how bottlenecks have shifted downstream to review, security, compliance, and deployment stages.
Context and Traceability Are the New Differentiators — The report explores why context and traceability are becoming critical advantages as code generation becomes commoditized. It examines barriers such as difficulty distinguishing AI-generated from human-written code, fragmented toolchains, and the lack of shared context across the software development lifecycle.
The Governance Imperative — This part analyzes the current state of governance for AI-generated code, including formal policies, human review practices, and the gap between adoption speed and structured controls. It shows that organizations with formal governance frameworks achieve better outcomes in human oversight and risk management.
The Four Pillars of AI Accountability — The report outlines four key pillars for success: integrated tooling that maintains context, traceability connecting code to original intent, governance paired with active human review, and long-term maintainability to prevent technical debt accumulation.
Supporting sections include key takeaways, methodology details, definitions of important terms (AI-generated code, code attribution, context, traceability, and governance), and practical guidance on building accountability at scale. The report concludes by emphasizing that the next phase of AI in software development will focus more on governing and maintaining code than simply generating it.
Overall, the document combines quantitative survey data, clear statistics, and actionable insights to help DevSecOps, security, and engineering leaders understand current realities and prepare for the governance challenges of agentic AI development.
Our Take
AI Governance Take
The GitLab AI Accountability Report makes it clear that governance has become the next critical frontier in AI-powered software development. While AI coding tools are delivering strong productivity gains and exceeding ROI expectations for most organizations, the acceleration of code generation has exposed significant gaps in context, traceability, and governance that many teams are still working to close.
The report shows that organizations are shifting from a focus on speed of code creation to the harder work of governing what happens to that code afterward. With 73% of DevSecOps professionals concerned about long-term maintainability and 84% identifying governance of AI-generated code as the biggest challenge, the findings underscore that sustainable AI adoption requires more than powerful coding assistants — it requires structured accountability systems.
For AI governance leaders, the key takeaway is the importance of building the four pillars outlined in the report: integrated tooling that maintains context, traceability that connects code to intent, governance frameworks with active human review, and processes that support long-term maintainability. Organizations with formal governance policies are already seeing better outcomes, including higher rates of human review and more structured risk management.
The report highlights that governance is not just a compliance checkbox but a strategic enabler. Teams that invest in traceability, clear ownership, and integrated platforms will be better positioned to scale AI safely, reduce technical debt, and maintain confidence in their software delivery process. Those that continue prioritizing speed without corresponding governance risk compounding technical debt and operational challenges.
Ultimately, the AI Accountability Report reinforces that the winners in the next phase of AI development will be those who master governance at scale. Context and traceability are becoming competitive differentiators, and organizations that treat AI-generated code with the same rigor as traditional development will gain a lasting advantage in both speed and reliability.
