Bye Bye Perspective API: Lessons for Measurement Infrastructure in NLP, CSS and LLM Evaluation

TL;DR

The paper discusses the impact of Perspective API's closure on NLP and LLM evaluation, highlighting issues of dependence on proprietary tools and advocating for open, reproducible measurement infrastructure.

Contribution

It analyzes the problems caused by reliance on a proprietary toxicity measurement tool and proposes requirements for an independent, open evaluation infrastructure.

Findings

Perspective's closure leaves non-updatable benchmarks and irreproducible results.

Dependence on proprietary tools causes epistemic issues in NLP research.

Calls for an open, adaptable toxicity measurement infrastructure.

Abstract

The closure of Perspective API at the end of 2026 discards what has functioned as the de facto standard for automated toxicity measurement in NLP, CSS, and LLM evaluation research. We document the structural dependence that the communities built on this single proprietary tool and discuss how this dependence caused epistemic problems that have affected - and will likely continue to affect - collective research efforts. Perspective's model was periodically updated without versioning or disclosure, its annotation structure reflected a single corporate operationalisation of a contested concept, and its scores were used simultaneously as an evaluation target and an evaluation standard. Its closure leaves behind non-updatable benchmarks, irreproducible results, and ultimately a field at risk of perpetuating these issues by turning to closed-source LLMs. We use Perspective's announced termination as an opportunity to call for an independent, valid, adaptable, and reproducible toxicity and hate speech measurement infrastructure, with the technical and governance requirements outlined in this paper.