Your developers shipped more code this quarter than last, and your scanner queue grew faster. The gap between those two numbers is a bill nobody has added up yet, and people who have never opened an AI coding assistant are already paying it.

The AI Remediation Bill, a study of 307 security, governance, platform, and engineering leaders at enterprises with 1,000 or more employees in North America and Western Europe, tracked both sides of that gap. On the productivity side, 93% report higher engineering output and 91% release more often. On the cost side, 94% say AI has increased the amount of software that needs remediation or cleanup, and 97% say their governance workload has gone up.

I work with security leaders and engineering teams on this every week, and the report matches what I see at the table.

One bill, five teams, no owner

The report's most useful finding is where the cost lands: in five budgets at once.

  • Engineering: 91% spend at least 5% of their time fixing or revisiting AI-assisted work. At a conservative floor, that is 6.7% of engineering time, or about 139 hours per engineer per year.
  • Operations: 76% had a release delayed by an unmanaged open source dependency in the past year, and 36% had more than one.
  • Compliance: 58% received an audit or compliance finding tied to open source governance.
  • Security: 32% confirmed a security incident traceable to an unmanaged dependency in the past two years.
  • Finance: 83% say the cost of managing open source risk is rising, and only 31% have ever quantified the total.

Each team sees its own line item and handles it locally. Engineering logs rework, operations reschedules a release, and compliance answers a finding, but nobody adds the charges together.

In governance reviews, the security leader usually brings an incident count and the platform lead brings a list of delayed releases. Neither realizes both numbers trace back to the same packages coming in the same way, so each team leaves with a new control for its own area.

92% have a policy. 52% follow it.

For CISOs, this is the number to bring to the next board conversation: 92% of organizations have a formal open source policy, and only 52% say it is consistently followed.

The gap predates AI, but AI made it expensive. When developers wrote every import by hand, a loosely followed policy leaked slowly. An AI coding assistant can add a package, its dependencies, and their dependencies in seconds, and 96% of respondents say AI has increased dependency growth. A policy that relies on a person pausing to check was written for a slower pace of development.

Half of respondents say a new dependency can reach production without any prior review. When I ask teams how they know what entered their environment last week, they usually point to the scanner report. A scanner report is a record of what already got in. Deciding what is allowed in has to happen earlier.

Governance that starts after the alert

Most of the governance the report describes happens after the fact. 64% say most of their governance activity follows an alert, and 63% address open source vulnerabilities only after a scanner finds them.

That explains why 55% of respondents named control, keeping unapproved open source from entering at all, as a bigger challenge than visibility (45%). Organizations have spent a decade buying visibility. Control is the piece they still lack.

The usual response is more policies, more tools, and more people, added one team at a time. Meanwhile, 83% say developer productivity is outpacing their governance capability, while 71% believe their current approach will scale. Expect that second number to fall as the first one rises.

The prevention gap

Respondents put 26% of their open source governance effort into prevention today and say it should be 37%. Among the organizations under the most strain, the gap is wider: 19% today, with 41% as the goal.

The people closest to the problem already know where their effort should go. What holds them back is the lack of a first step that works without a new budget line or agreement across five teams.

Software supply chain security is a sourcing problem before it is a scanning problem. Each hour moved to the point where packages enter your environment saves work for every team downstream, because the same charge stops recurring in five places.

Where to start this quarter

The organizations I see make the most progress begin by deciding the order of the work.

  1. Put one number on the bill. Pick a single quarter. Ask engineering for rework hours, operations for delayed releases, compliance for findings, and security for incidents tied to open source. Put them on one page. Only 31% of organizations have done this, and it changes every budget conversation that follows.
  2. Name one owner for intake. Five teams pay the bill, and none of them decides what comes in. Give one person the authority to decide which packages are eligible to enter. They don't have to do every review themselves; they own the rule.
  3. Govern AI-introduced dependencies first. Start where the volume is growing fastest. Require packages pulled in by AI coding tools to come from an approved, governed source before they reach a build, and track how many arrive each week.
  4. Require provenance evidence from suppliers. For every component, ask where it was built, from what source, and with what attestation. SLSA gives security and engineering a shared standard for provenance that they can hold any supplier to.
  5. Make the scanner your second check. Keep scanning. When the packages entering your environment are already governed, the scanner verifies what you allowed in, and the alert queue shrinks for every team.

Each of these steps can begin with a meeting you schedule this week.

A smaller bill

Learning that an AI investment generates costs in five places at once is hard news.

The teams I've seen take the first steps (one owner, one number, and one governed source for AI-introduced packages) see rework and release delays fall sooner than they expect. Most of the bill comes from letting packages arrive ungoverned, and an organization can change that decision this quarter.

Your developers will keep getting faster. Governing open source at the point of origin lets every team downstream keep pace with them.

The full findings, including the cost breakdown by team, are in The AI Remediation Bill.

About the Author: Leslie Pascual is Field Engineering Manager, AI & Security Solutions at ActiveState. Leslie Pascual works with security leaders and engineering teams to govern the open source their software is built on. She translates between what executives assume about software supply chain risk and what engineering teams deal with day to day.

Leslie Pascual — Field Engineering Manager at ActiveState https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhbqVyOlJvFQ9TXXQhkHUO3qa6OozwKfWd7vXtK-7-qqMKL4YHDMybD1ynNa7fPrMC71irT_p5mg89qDTAks4MVcM2LUMKEFL-G897pvTAhU2uMonXM-BblXbEBXOHTzMx_XbDzVDmYCbiPcs9uJyhazgQzP1MBZCkXtXTb7yrJ3yx9OapUh4i4CML5tCQ/s1700-e365/Leslie.png
Found this article interesting? This article is a contributed piece from one of our valued partners. Follow us on Twitter and LinkedIn to read more exclusive content we post.