[ad_1]
Enterprises Are Giving AI Agents Cloud Access Despite Near-Universal Concerns, Survey Finds
Enterprise software teams are giving artificial intelligence agents access to cloud environments despite widespread concerns about leaked credentials, misconfigurations, and costs, according to a new survey from LocalStack.
The State of AI in Cloud Development 2026 report found that 79% of respondents grant agents cloud access for development, testing, or staging, while 97% expressed concerns about providing that access. Among those granting access, 26% said agents can also reach production environments.
Dimensional Research surveyed 309 hands-on developers and team managers at organizations with more than 1,000 employees in the United States and Canada. The online survey captures respondents’ experiences and estimates, including how much AI-generated code their teams actually ship.
LocalStack sells a local cloud development platform for developers and AI agents.
Note that LocalStack has a direct connection to the AWS development ecosystem: Its LocalStack for AWS product emulates AWS services on local infrastructure so teams can develop and test applications before deploying them to the cloud, although the findings cover public-cloud development generally rather than AWS alone.
Access Extends Beyond Development
Of all respondents, 61% said agents occasionally have access to cloud development, testing, or staging environments, and 18% said they always do. The remaining 21% said agents never have that access.
The leading concern was the potential for leaked credentials, cited by 67%, followed by misconfigurations at 62%, increased cloud costs at 53%, and failures in production at 51%. Just 3% reported no concerns.
Concerns also persisted among respondents who always grant agents cloud access: 92% of that group expressed reservations. Misconfigurations led their list at 65%, followed by potential credential leaks at 55%.
Among respondents who give agents cloud access, 89% allow access to testing environments, 85% to development, 55% to staging, and 26% to production. Those percentages describe the access-granting subgroup, rather than the full survey population.
Access was more common at larger employers. Among respondents at organizations with more than 5,000 employees, 84% granted agents cloud access, compared with 72% at organizations with 1,000 to 5,000 employees.
More Code Does Not Mean More Releases
AI has increased code output for many surveyed teams: 65% said they generate at least twice as much code as before adopting AI, and 32% reported at least a fivefold increase. But respondents estimated that only 61% of AI-generated code reaches production, on average.
[ad_2]
Source link
Leave a Reply