Distributed Attacks in Persistent-State AI Control: Introducing Iterative VibeCoding

Paper Overview

• Research area: cs.AI
• Authors: Josh Hills, Ida Caspary, Asa Cooper Stickland
• Published: 2026-07-02
• arXiv: 2607.02514

Abstract

As AI coding agents become more autonomous, they increasingly ship code iteratively, with the codebase persisting across sessions. This persistence creates a new attack surface: a misaligned or prompt-injected agent can distribute attacks across pull requests (PRs) and time its payload for the PR with the best natural cover. The authors introduce Iterative VibeCoding, a setting for AI control research that captures this threat model.

Why It Matters

• Persistent codebases across agent sessions enable multi-step, distributed attacks rather than single-shot exploits.
• Timing an attack to coincide with naturally noisy or risky-looking PRs complicates monitoring and detection.
• Iterative VibeCoding provides a benchmark-style framing for evaluating AI control methods under realistic, iterative development workflows.

Paper link: https://arxiv.org/abs/2607.02514