AI Is Breaking Open Source Trust — And Provenance Is the Fix

/ AI, Application Development, Open Source

The News

On a recent episode of the AppDevANGLE podcast, host Paul Nashawaty spoke with Timothy Lewis, co-founder and CEO of tea.inc., a company focused on open source security, package provenance, and trust infrastructure for software supply chains. Lewis, a 40-year veteran of software engineering with deep roots in cryptography and open source, argued that AI-generated code is fundamentally altering the threat model for open-source package ecosystems, overwhelming maintainers with pull requests, degrading provenance integrity, and outpacing the governance structures enterprises rely on. The conversation covered maintainer economics, cryptographic signing gaps, SBOM adoption failures, and what a credible provenance and identity foundation for open source would actually require.

Analyst Take

The real supply chain risk isn’t the code. It’s the economics behind it.

The framing Lewis brings to this conversation is worth taking seriously. Most supply chain security discussions fixate on tooling: vulnerability scanners, SBOM generators, runtime protection. Lewis points to something more structural. Open source maintainers, many of them unfunded individuals, are now operating as critical infrastructure for global enterprises while absorbing an AI-generated flood of pull requests. Contribution volume has decoupled from review capacity. That gap is a security gap, and no amount of policy enforcement at deploy time closes it if the upstream packages were already compromised or abandoned.

This is not a theoretical concern. Lewis cites that only six to seven percent of open source software was signed before the current wave of AI-accelerated development. That baseline is indefensible given what organizations are now building on top of it. When you pair that with the observation that only 1.6% of organizations have adopted SBOM requirements in response to supply chain attacks, the picture is clear: governance is lagging threat exposure by a wide margin. ECI Research’s own 2026 Application Development survey found that 29.1% of respondents selected “AI-generated package risk” as their biggest open-source security concern in 2026, making it the top-ranked concern ahead of malicious package injection and transitive dependency sprawl. The market has named the problem, but it has not yet funded the solution.

Why provenance is the missing primitive

Lewis’s argument about cryptographic signing is the most technically important point in the conversation. The problem isn’t that signing technology doesn’t exist; it’s that there was never a strong incentive to adopt it, and common practices like squash merges have actively eroded the signature history that does exist. The result is a software ecosystem where you can trace a package to a repository, but not reliably to an identity, a history of contributors, or a chain of custody across VCS migrations.

For developers, this matters architecturally. The dependency graphs that underpin modern applications assume trustworthy upstream packages. If provenance is opaque, the entire trust model of the build pipeline rests on assumption rather than verification. Lewis’s framing of this as an “evidence indexing problem” is precise: what’s needed isn’t just signing at the point of release, but a continuously maintained, cryptographically verifiable index of contributor identity, build history, and dependency lineage across the full lifecycle of a project. That’s an expensive infrastructure problem, and no one is adequately funding it yet.

The governance gap ITDMs should not ignore

For IT decision makers, the headline is straightforward: your security posture is only as strong as the weakest unsigned package in your dependency tree. The 5.6% governance adoption figure Nashawaty cites in the conversation suggests that most organizations are responding to supply chain threats reactively, if at all. ECI Research’s 2026 Application Development survey reinforces this: 47.4% of respondents selected “Software supply chain security” as a top investment priority for the next 12 months, yet actual adoption of structural controls like SBOMs and provenance tracking remains thin. The gap between stated priority and funded action is where breaches happen.

Lewis also makes a pointed observation about regulatory momentum. NIST frameworks are beginning to require deeper evidence of provenance validation, and he believes enforcement pressure, whether through penalties or procurement requirements, could finally move the needle on signing adoption. Organizations in financial services and healthcare are already feeling this: ECI Research’s 2026 Application Development: Day 1 survey found that 71.5% of respondents selected “Industry-specific compliance (FinServ/Healthcare)” as a regulatory pressure influencing release engineering. That pressure is a forcing function. Companies that wait for a breach or a regulatory penalty to operationalize provenance controls will pay more, in both cost and credibility, than those who build the capability now.

Looking Ahead

The conversation with Lewis points to a consolidation moment in open source security infrastructure. The tooling layer, scanning, signing, and SBOM generation, is maturing. What’s missing is the economic layer: reliable, scalable mechanisms to fund maintainers, distribute value based on actual dependency usage, and sustain the evidence-indexing infrastructure that enterprise-grade provenance requires. Expect to see commercial platforms emerge that bundle provenance attestation with dependency intelligence and compliance reporting, targeting the enterprise procurement workflows that Lewis describes. tea.inc. is one early entrant in this space; others will follow as regulatory pressure converts “nice to have” into contract requirement.

The deeper shift is that software supply chain security is moving from a DevSecOps concern into a board-level risk category. As AI accelerates both the generation of code and the sophistication of supply chain attacks, the organizations that treat provenance as a first-class architectural requirement, not an audit checkbox, will hold a structural advantage. The economics of open source are being renegotiated right now. ITDMs who engage with that renegotiation proactively, by requiring signed packages, funding upstream maintainers, and implementing SBOM workflows, will be better positioned than those who wait for the ecosystem to solve it for them.

Author