Brief IA
›

AI Code: 81% of Teams Merged Without Fully Understanding

Business & Startups·Tom Levy·

AI Code: 81% of Teams Merged Without Fully Understanding

AI Code: 81% of Teams Merged Without Fully Understanding
Key Takeaways
1According to TestMu AI, 81% of the teams surveyed have merged AI-generated code without fully understanding it.
2Only 13% have an incident response procedure for errors produced by AI.
3Automated checks before merging are implemented by 41% of the teams.
Why it matters : Verification practices are not keeping pace with the volume of generated code, according to the leader of TestMu AI.

According to TestMu AI, 81% of surveyed teams admit to merging AI-generated code without fully understanding it. Its State of AI in Quality Engineering 2026 report describes disparate control practices, while its leader recommends independent validation of the producing model.

Only 13% of teams prepare for incident response

According to TestMu AI, only 13% of teams have an incident response mechanism for already published AI-origin errors. Asad Khan, co-founder and CEO of TestMu AI, advocates for subjecting any AI production to an independent validation system of the model from which it originates. This control should occur before any merge and then after each modification.

The leader claims that the teams surveyed applying this discipline experience fewer defects in production. He predicts that this practice will become the norm among engineering organizations delivering code with AI.

La lettre du soirGratuit · 19 h

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

A second model does not guarantee understanding

Some teams delegate code evaluation to another large language model. For the shipment of misunderstood code, the survey indicates a rate of 29% among teams using a model as a judge, compared to 15% among those not employing this method.

According to Asad Khan, the use of AI now encompasses both code writing and test production. Teams are also increasingly entrusting it with assessing their readiness for delivery.

Automated checks concern 41% of teams

Before integrating generated code into an existing base, often already in production, 41% of teams perform automated checks, according to TestMu AI. Human review is practiced by 55% of teams; 51% involve a dedicated quality assurance team.

Asad Khan believes that verification capabilities are progressing slower than the volume of code produced.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.