preloader

· · devops cloud ai security devsecops

99 Percent of Organizations Let AI Agents Act in Development Pipelines, Cloud Security Report Finds

Source: Palo Alto Networks

Source: Palo Alto Networks

Palo Alto Networks released its sixth annual State of Cloud Security Report on 16 November 2026, titled Beyond Human Scale: How AI Agents Reshape Cloud Risk. The report is based on a survey of 2,800 security leaders and practitioners across 10 countries and 13 industries, and it documents a widening gap between the speed of AI-driven development and the ability of security teams to keep up.

Agents run the pipeline, security runs behind

The headline finding is that 99 percent of organizations rely on AI agents for autonomous or semiautonomous actions in their development pipelines, and 82 percent report faster cloud infrastructure update cycles as a result. Nearly three-quarters, 73 percent, allow AI agents to modify code, pipeline configurations or production environments. Yet on average organizations enforce fewer than three of the eight pre-deployment security checks evaluated in the research, and respondents report that 19 percent of high and critical severity security issues reach production during a typical release cycle. Business pressure to accelerate releases and unclear risk acceptance policies remain the main barriers.

Identity and visibility gaps

Autonomous agents also create an identity problem. 40 percent of organizations cannot fully reconstruct what an AI agent did after it executed, and 53 percent log agent activity under human developer identities or personal access tokens, making it impossible to tell human actions from machine actions during an incident. Credential hygiene lags too: organizations take an average of 2.5 days to revoke or rotate compromised credentials, and 60 percent still do it manually. On the data side, only 9 percent can automatically discover ungoverned AI use in their cloud environments, while 40 percent have found sensitive governed data copied into ungoverned AI tools and 39 percent have found regulated data used for unauthorized model training or fine tuning.

Meanwhile the offensive side is accelerating. 65 percent of security leaders consider frontier AI-powered autonomous vulnerability discovery and exploitation a severe threat to critical infrastructure within the next 12 to 18 months, while organizations still take an average of 18.4 days to move from detection to response.

What to do about it

The report points to where DevSecOps effort should go in 2027: give every agent its own non-human identity with scoped, short-lived credentials, enforce pre-deployment controls in the pipeline itself rather than after it, generate audit trails that distinguish agent actions from human ones, and automate credential rotation so response no longer depends on someone’s weekend. European teams have an extra reason to move first: NIS2 makes management bodies personally accountable when exactly this kind of unmonitored change leads to an incident.

If AI agents already touch your pipelines and you want that reviewed by people who do this daily, contact Excello Digital and we will help you build agent governance that keeps the speed without the blind spots.

These news items are automatically aggregated from industry sources and are not individually reviewed. Any inaccuracies are unintentional — let us know and we'll correct or remove it.

We’ll help you resolve your infrastructure challenges

Our team of experts is ready to help you with your infrastructure challenges. We’ll give you honest and personal treatment. Get in touch to learn more.

Get in touch!