BlueFlag Security, the Developer Risk and Governance Platform, today announced expanded AI Agent Governance that fully maps the access, behavior and oversight of both approved and shadow AI agents, including options for automated risk remediation. Activated within BlueFlag’s pioneering platform, customers can now govern all aspects of AI agents and their relationships with human/machine identities, tools, pipelines and production applications across the software development lifecycle (SDLC). The new capabilities follow the release of BlueFlag’s 2026 State of Developer Identity Risk, a first-of-its kind research report that provides evidence of how little visibility exists in some major organizations around massive SDLC risks that are invisible to code scanning tools and enterprise identity management platforms.
AI has created a tectonic shift in software development. No longer only coding assistants, AI Agents are empowered by developers to independently plan the work, delegate to other agents, and provision the environments they need, and access on the developer’s behalf, and with developer access. BlueFlag’s new data reveals that 83% of organizations run AI coding agents no one formally approved, and 56% of organizations have autonomous agents operating as their own identity. Among early adopters, AI now writes up to 72% of all commits. Where agents work under a developer’s identity, most of those identities hold administrator permissions and are not monitored at all, and in more than 1 in 5 organizations an autonomous agent has already merged code with no human in the loop.
“An AI agent is now the most powerful and least governed identity in the development environment,” said Raj Mallempati, CEO of BlueFlag Security. “That level of trust is at best misguided and at worst a potentially catastrophic threat to the business. BlueFlag offers the unparalleled ability to govern agents where they work, alongside the human developers and non-human identities in the same ecosystem, and connect their behavior to everything around them, so an attack path shows up the moment one manifests. Agents are about to become the primary developers of software. People will set the goals, the guardrails, and where a human has to sign off. BlueFlag is already the governance layer across both, and early next year it will begin to interrupt an attack path as it forms, not only report it.”
The expansion established 6 new and/or enhanced capabilities that bridge agent identification, governance and risk remediation. Regardless of human, machine or AI, BlueFlag’s groundbreaking platform discerns identity type, measures access and permissions against actual use, baselines behavior, and alerts when that behavior changes. The new agentic AI capabilities include:
-
Identity Discovery and Classification
- AI Agent Classification and Approval: BlueFlag discovers every agent from its activity in code changes and builds and classifies it by whether it writes, tests, reviews, and/or deploys code, and whether it works under a developer’s identity or operates as its own. It shows the developers who work with each agent, the repositories and pipelines it operates in, and every commit and pull request it has written, reviewed, or approved. It also marks which agents the organization approved and which were not.
- Shadow AI on Developer Devices: For AI developed code created outside of the repository, BlueFlag integrates with endpoint detection and response (EDR) platforms, including Crowdstrike and SentinelOne, to identify AI tools running on developer devices. This does not require new software on the endpoint.
-
AI Oversight and ROI
- Human-in-the-Loop Checks: BlueFlag flags pull requests an AI agent approved, commits written entirely by an agent, and the pipeline runs that were deployed without human review, tracing each from the agent to the build. BlueFlag also can require a human review, at pull request before any merge or deployment. For any release, the organization can document and show where a person was in the decision, and where one was not.
- Agentic Application Inventory: The AI tools developers use to write code are only half the story BlueFlag also identifies every application in the codebase that uses AI and which model providers it calls, which agent frameworks it is built on, and whether the model runs on the company’s own infrastructure or a third party. Finally, it measures each against the OWASP Agentic Top 10 to prioritize any corrective action.
- AI Spend Against Developer Output: AI budgets are growing faster than anyone can show a return on them. BlueFlag matches the tokens consumed and the dollars spent, to the work produced with them: the commits, pull requests, and accepted lines of code, by developer, tool, and model. This includes how much of the codebase AI wrote and what the AI spend returned, ROI based on each model in use, which paid seats are idle, and where spend is rising with no commits to show.
-
Flexible Automated Action
- Responsible and Responsive Remediation: BlueFlag has also introduced 2 new remediation features that let customers address every risk in one of 3 ways: step-by-step manual instructions, a generated remediation script, or a one-click automated fix that can be rolled back. AccessGuard remediates identity hygiene, from over-permissioned identities to stale credentials and dormant team memberships. DriftGuard remediates misconfigurations across repositories and pipelines, from missing branch protections to the other settings that drift over time.
“We accept some unknowns when the attack comes from outside. We should not have to accept them in our own development environment, and agentic AI has made that gap bigger overnight,” said Kristie Pfosi, Global Chief Information Security Officer at Marelli. “Within the first week BlueFlag showed us which AI agents were working in our repositories, which developers they worked with, and where the gaps were in how they were approved, scoped and reviewed. Those are findings we can act on the same day, and no other tool we run could have surfaced them.”
This new platform expansion also coincides with BlueFlag’s October brand relaunch, which returns the company to its namesake, the “Blue Flag” awarded to beaches that meet a strict standard for the water beneath the surface. As part of this, BlueFlag also introduced the brand film, “Beneath the Surface,” a rebuilt website, and an interactive brand-story page.
The 6 capabilities are generally available now. BlueFlag connects through the APIs of the tools customers already run, read-only by default, and is live in under 5 minutes with nothing installed and nothing in the pipeline. Automated fixes run only when the customer chooses them, and BlueFlag never touches the code.code. As AI agents take on more of the software development lifecycle, BlueFlag will keep extending the platform so that the people responsible for the software stay in control of it. Get a demo at blueflagsecurity.com.
About BlueFlag Security
BlueFlag Security is the Developer Risk and Governance Platform for the software development lifecycle (SDLC). It gives security and engineering teams one view of who and what acts on their code: human developers, non-human identities, and now AI coding agents, the most powerful identities in the environment, together with the tools and pipelines they work through. BlueFlag maps what each identity can reach and what it is permitted to do, learns how it normally behaves, alerts when that changes, and connects the signals when they form an attack path. Every finding comes with its fix, guided or automated. BlueFlag connects through read-only APIs in minutes, with nothing installed, and never touches the code. Founded in 2024 and headquartered in Milpitas, California, the company has raised $28 million from Maverick Ventures and Ten Eleven Ventures. Learn more at blueflagsecurity.com.
View source version on businesswire.com: https://www.businesswire.com/news/home/20261104963218/en/
Media gallery
