# New Paper Tests General-Purpose Language Models Against Darknet CAPTCHAs

The challenges are built to work without JavaScript, which gives them visual traits that defenders assumed would require attackers to train bespoke models.

- Published: 2026-09-01T06:29:06.347Z
- Canonical: https://polylog.news/ai/2026-09-01/new-paper-tests-general-purpose-language-models-against-dark
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.CR](https://arxiv.org/abs/2608.28794), [Anthropic Research](https://www.anthropic.com/research/team/frontier-red-team)

A paper posted to arXiv evaluates automated methods for solving CAPTCHA challenges found on darknet sites, using general-purpose large language models rather than task-specific classifiers. Because those sites must work for visitors who dis…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-09-01/new-paper-tests-general-purpose-language-models-against-dark (subscription information: https://polylog.news/pricing).