GitLab’s 2026 AI Accountability Report highlights: AI paradox: Although 78% of developers say they code faster, overall software delivery has not accelerated due to bottlenecks in downstream testing and verification, as well as new governance and traceability challenges.
According to GitLab research, AI has made writing software faster: 78% of respondents reported faster code output and 73% said overall code quality improved. However, AI tools have uncovered a deeper problem: companies cannot easily control what they ship because governance, traceability and accountability cannot keep up, leading to a structural imbalance.
The report defines AI accountability as the organizational and technical ability to answer three questions about each line of AI-generated code: where does it come from, what should it do, and who is responsible for it once it is in production? Most organizations today cannot answer these questions.
In fact, 85% of respondents “agree that AI has shifted the bottleneck from writing code to reviewing and validating it.” As a result, 79% say the overall software delivery process hasn’t kept up with coding.
As Manav Khurana, chief product and marketing officer at GitLab, notes, recent events such as supply chain attacks, reliability issues and regulatory expectations show that traceability is a critical concern to prevent business compromise. Respondents point to three main factors that contribute to making traceability difficult: difficulty distinguishing between AI-generated and human-written code (43%), fragmented toolchains (40%), and systems that do not trace code origin (39%). GitLab’s report reflects this gap, noting that:
87% are confident their team can determine whether AI-generated code contributed to a production incident within 24 hours. [only] 34% of organizations that experienced an incident last year were unable to make this determination.
For 85% of respondents, the solution lies in stronger governance, i.e. establishing clear guidelines to ensure the provenance and accountability of AI-generated code. Without it, 83% of companies view the accumulation of AI-generated code as a risk, and 44% rank it among their top technology concerns.
The results of the GitLab research echo the views of a previous Reddit thread, in which the OP notes that continued investment in AI increased “speed at the text editor/terminal level,” but they spent most of their time “wading through the quicksand of Agile/Jira and middle management bloat.” Another user, YourMatt, similarly noted that while advances in coding speed were impressive, they did little to address the broader inefficiencies that ultimately limit deployment:
Sprint after sprint, however, no one in our focus group produced more story points than before. It really highlighted that the mechanics of coding are a relatively small part of our work.
In a recent thread, Mestyo reiterates this view and argues that the majority of work done by individual contributors cannot be meaningfully accelerated by AI coding tools.
As a final note from the community, Reddit users EveryDay_is_LegDay takes this view, arguing from experience that testing remains the biggest bottleneck and that “faster code production only exacerbates the problems of most development teams.”
https://www.infoq.com/news/2026/06/ai-coding-outpaces-governance/
