Safe Code Warrior has launched an AI Adoption Mannequin for software program improvement, geared toward Chief Data Safety Officers managing AI use throughout improvement groups.
The framework outlines three phases of AI use in software program improvement: AI-Assisted, AI Native and Agentic. Every stage is tied to totally different ranges of danger, developer coaching wants and governance controls as organisations broaden their use of AI instruments and autonomous techniques in coding workflows.
The launch comes as firms face rising stress to control AI use in software program engineering past conventional developer groups. Staff utilizing no-code instruments and so-called vibe coding approaches also can add to an organisation’s software program danger profile, even when they don’t seem to be skilled engineers.
Analysis cited from Gartner’s 2026 Hype Cycle for Safe Software program Engineering says AI-augmented improvement is “increasing the assault floor sooner than conventional controls can scale”, whereas AI coding instruments are making safe coding abilities extra vital.
Safe Code Warrior is positioning the framework as a manner for safety leaders to evaluate present AI adoption, align coaching to that stage and determine which controls must be in place. The construction is meant to assist safety groups monitor danger amid what the corporate describes as a shift from the Software program Growth Lifecycle to an Agentic Growth Lifecycle.
Three phases
Within the first stage, AI-Assisted, builders use AI in a restricted supporting position. The subsequent stage, AI Native, displays deeper integration of AI into improvement work, whereas the ultimate Agentic part covers extra autonomous orchestration of software program duties.
The framework is meant to provide CISOs a sensible place to begin relatively than treating AI use as a single class. That distinction issues as a result of organisations are adopting AI inconsistently, with some groups experimenting cautiously and others transferring in direction of techniques that may perform broader improvement features with much less human intervention.
For safety leaders, one problem is linking the unfold of AI coding instruments to measurable controls and spending choices. The framework is designed to assist data-led choices on danger, coaching and governance, particularly as boards and know-how leaders face nearer scrutiny over the associated fee and oversight of AI tasks.
One other problem is the failure fee of tasks that lack controls. Gartner has predicted that by 2027 greater than 40% of agentic AI tasks can be deserted due to uncontrolled prices and poor danger controls, in response to figures cited by the corporate.
Safety abilities
Safe Code Warrior argues that the rise of AI in improvement modifications the developer’s position relatively than eradicating duty for software program safety. Meaning firms have to rethink coaching as AI instruments develop into extra widespread in day by day coding work.
“In our present AI-powered improvement, writing traces of code is nearly free, however builders are nonetheless on the hook for safe outcomes. Their safety abilities have to evolve from code author to creator & orchestrator,” mentioned Pieter Danhieux, Co-founder & Chief Government Officer, Safe Code Warrior.
One of many mannequin’s important makes use of is to tailor coaching to how particular person groups use AI. Somewhat than making use of the identical instruction throughout an organisation, companies ought to map abilities and studying must the stage of adoption and the diploma of autonomy concerned in improvement duties.
That strategy displays a broader debate within the software program sector over whether or not automated instruments must be managed primarily via technical safeguards or via stronger schooling for the folks directing them. Safe Code Warrior argues that coaching builders to make use of AI accurately from the outset is a simpler technique to scale back repeated vulnerabilities and handle prices than relying solely on extra layers of AI oversight.
Danhieux mentioned the framework was constructed to deal with that governance problem as software program improvement strategies change.
“CISOs want an strategy to ADLC governance that’s as trendy because the methodology itself, one which follows an adoption mannequin designed for agentic AI’s evolving, adaptive strategy to software program improvement. We have constructed this framework to assist organisations flip safe AI adoption and AI governance from a reactive train right into a measurable, scalable self-discipline,” mentioned Danhieux.
The AI Adoption Mannequin is now accessible as organisations search clearer methods to measure the place AI is being utilized in software program improvement and what controls are wanted as that use turns into extra autonomous.