AI code destabilizes deploys
A new Harness survey flagged that over 50% of respondents say AI-generated code causes deployment problems at least half the time — a clear red flag for teams shipping ML-enabled features. devopsdotcom.
Harness’s State of DevOps Modernization surveyed 700 engineers and technical managers across the U.S., U.K., France, Germany and India harness.io, and the fieldwork was carried out by Coleman Parkes on Harness’s behalf. devops.com The study found 84% of respondents use AI tools daily for coding devops.com, while 78% admitted their delivery toolchains are fragmented and 70% reported pipelines plagued by flaky tests and deployment failures. devops.com Teams that use AI coding “very frequently” showed sharper downstream strain: 45% deploy to production daily or faster and 22% of those deployments resulted in a rollback, hotfix, or customer‑impacting incident, with mean time to restore at about 7.6 hours for very‑frequent users. harness.io Security and governance gaps stood out: a prior Harness study reported 48% of leaders expect AI to increase vulnerabilities and only 41% were fully confident governance would catch issues before production, while 72% said their orgs had already suffered at least one production incident caused by AI‑generated code. prnewswire.com
