Three in Four Companies Have Suffered an Outage Caused by AI Code

Three out of four companies worldwide have already experienced a production outage to which code written by artificial intelligence, an AI agent, or an AI-based tool contributed. That is the finding of a Futurum Group survey of 839 IT managers worldwide, published as part of a market report on software lifecycle engineering for the second half of 2026.

The study, first reported in Poland by CRN Polska, was produced as a companion piece to Futurum Group's market forecast for the Software Lifecycle Engineering segment. Analysts estimate that the market for tools supporting the entire software lifecycle, currently worth $111.1 billion, will grow to $226 billion by 2030, at a compound annual growth rate of 15.3 percent. The reason: artificial intelligence has stopped being merely a developer's assistant and is increasingly driving most of the work on code on its own.

The scale of the problem

The survey shows that AI adoption in software development has outpaced oversight of it by years. 75 percent of respondents admitted that their organization had a production incident in which code generated by AI, an autonomous coding agent, or another AI-based tool was one of the contributing factors. What's more, 42 percent of companies reported that such events recurred several times over the course of a year, and only 7 percent of surveyed managers were certain that nothing similar had happened at their company.

The report's authors stress that these are not isolated cases of minor bugs. At 54 percent of the surveyed organizations, AI already accounts for more than half of the entire software development lifecycle, and 40 percent of companies admit that artificial intelligence generates most of the code that ultimately reaches production. The scale of deployment is growing faster than the quality and security controls meant to oversee that code.

A gap in oversight

The most troubling finding is the gap between the pace of adoption and the maturity of governance processes. Only 43 percent of companies require AI-generated code to be verified by a human before deployment. Governance of coding agents was identified as the least mature engineering practice in the entire survey, with only 18 percent of organizations having standardized or fully mastered processes in this area.

The market has already priced in AI taking over code generation, but it's underestimating the bill for overseeing it - Mitch Ashley, Vice President and Practice Lead for Software Lifecycle Engineering at Futurum Group

The title of Futurum Group's report itself is telling: companies handed artificial intelligence the keyboard before they had time to build guardrails. According to the analysts, it is precisely this sequence, scale first, control second, that explains the rising number of incidents reported by IT managers around the world.

Not the first such signal

Futurum Group's data is not an isolated voice. Earlier this year, Palo Alto Networks published its State of Cloud Security 2025 report, based on a survey of more than 2,800 security professionals across ten countries. It found that 99 percent of teams already use generative AI when writing code, and 99 percent of organizations experienced at least one attack targeting AI-based systems in the past year. More than half of companies deploy new code weekly or more often, but only 18 percent of security teams can patch vulnerabilities at a comparable pace.

Most organizations allow the use of generative AI tools for writing code. At the same time, only a few have conducted a formal risk assessment - Wojciech Gołębiowski, Palo Alto Networks Central and Eastern Europe

What it means for companies in Poland

For Polish companies increasingly eager to adopt AI-assisted coding tools, both studies signal that simply deciding to let such tools into development teams' workflows is not enough. Without a formal policy, mandatory human code review, and control over the permissions granted to agents, the risk of pushing errors into the production environment grows in proportion to the pace at which AI takes over writing code.

Futurum Group's report also fits into a broader trend of rising enterprise spending on quality control for AI-generated code. Companies are starting to buy dedicated tools for scanning and reviewing such code, but as the numbers show, demand for oversight still isn't keeping pace with how quickly AI is moving into production.