Most Enterprises Now Give AI Agents Cloud Access

A 309-respondent survey finds leaked credentials are the top worry as agents reach development and staging systems today.

Updated on Nov. 4, 2026 in Artificial Intelligence

Long server aisle in a modern data center with dark steel racks receding toward a central vanishing point under cool blue light, illustrating enterprise cloud infrastructure. — Survey of 309 developers finds 79% of large U.S. and Canadian organizations give AI agents access to cloud development, testing, or staging environments. AI Illustration. Upload story photo >

Live Poll

Would you be comfortable letting an AI tool access your important accounts and data?

A LocalStack survey of 309 developers and team managers at large U.S. and Canadian organizations found that 79% grant AI agents access to cloud development, testing, or staging environments. This is a survey finding about current practice, not a product or standard; it describes how enterprises are deploying agents today.

Why it matters

Enterprises are extending autonomous AI agents into cloud infrastructure faster than security controls are settling, which makes access governance a central question for agent deployment. The concerns respondents raised, leaked credentials and misconfiguration, are the failure modes that will shape how agent permissions are designed.

The survey covered 309 developers and team managers at organizations with more than 1,000 employees in the United States and Canada, with 61% reporting that agents occasionally have cloud access and 18% reporting always-on access. Respondents named leaked credentials (67%), misconfigurations (62%), increased cloud costs (53%), and production failures (51%) as concerns, but the survey does not report what share of agents have already caused any of these problems.

The players

LocalStack

LocalStack sells a local cloud development platform used by developers and AI agents to test cloud applications without live infrastructure.

Dimensional Research

Dimensional Research is a market research firm that conducted the online survey of 309 developers and team managers for LocalStack.

The details

Developers increasingly let AI agents act on their behalf inside cloud accounts, where agents can create, modify, and test resources. Cloud development, testing, and staging environments are the typical first step, because mistakes there are assumed to be cheaper than in production. The survey's figures suggest that boundary is porous: 26% of respondents who grant access say agents can reach production systems. Credentials are the core exposure, since an agent holding a leaked key inherits whatever permissions that key carries.

November 4, 2026: LocalStack's survey findings were published on Campus Technology.

The Tech Race

The survey places enterprise agent adoption against the cloud-development tooling that agents rely on, with LocalStack's own platform among the systems being used this way. Competing efforts to govern agent permissions in cloud accounts are not covered by the survey, so the race to set access standards remains unsettled.

Developers at large organizations already using agents in cloud accounts face the access-control decisions the survey describes, and teams that have not yet set agent permissions can review which environments their agents reach today. The survey does not report pricing or platform changes, so the practical near-term effect is limited to internal policy on credentials and environment access.

The takeaway

Agent access to cloud environments is now the default at most large enterprises, with production access a minority but real practice. Watch whether LocalStack or Dimensional Research publishes follow-up data on agent-caused incidents, since that would show whether the 67% credential concern is turning into measured harm.