# Careless Whisper: Speech-to-Text Hallucination Harms arXiv: https://arxiv.org/abs/2402.08021 ยท HTML: https://mlsanity.com/paper/2402.08021/ ## Key facts - Cited by 53 papers in the map (data through 14 Oct 2026). - Builds on 8 of its 45 references that are in the map. - First posted on arXiv 12 Feb 2024. ## Builds on (most cited first, 8 of 8) - [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022) - [Survey of Hallucination in Natural Language Generation](https://mlsanity.com/paper/2202.03629/) (2022) - [Fairness in Machine Learning](https://mlsanity.com/paper/2012.15816/) (2020) - [Powerset multi-class cross entropy loss for neural speaker diarization](https://mlsanity.com/paper/2310.13025/) (2023) - [End-to-end speaker segmentation for overlap-aware resegmentation](https://mlsanity.com/paper/2104.04045/) (2021) - [Hallucinations in Neural Automatic Speech Recognition: Identifying Errors and Hallucinatory Models](https://mlsanity.com/paper/2401.01572/) (2024) - [Augmented Datasheets for Speech Datasets and Ethical Decision-Making](https://mlsanity.com/paper/2305.04672/) (2023) - [Hallucination of speech recognition errors with sequence to sequence learning](https://mlsanity.com/paper/2103.12258/) (2021) ## Built on by (most cited first, 30 of 53) - [CrisperWhisper: Accurate Timestamps on Verbatim Speech Transcriptions](https://mlsanity.com/paper/2408.16589/) (2024) - [Investigation of Whisper ASR Hallucinations Induced by Non-Speech Audio](https://mlsanity.com/paper/2501.11378/) (2025) - [Lost in Transcription, Found in Distribution Shift: Demystifying Hallucination in Speech Foundation Models](https://mlsanity.com/paper/2502.12414/) (2025) - [Anatomy of Industrial Scale Multilingual ASR](https://mlsanity.com/paper/2404.09841/) (2024) - [Calm-Whisper: Reduce Whisper Hallucination On Non-Speech By Calming Crazy Heads Down](https://mlsanity.com/paper/2505.12969/) (2025) - [Provocations from the Humanities for Generative AI Research](https://mlsanity.com/paper/2502.19190/) (2025) - [LLM-based speaker diarization correction: A generalizable approach](https://mlsanity.com/paper/2406.04927/) (2024) - [Hallucination Benchmark for Speech Foundation Models](https://mlsanity.com/paper/2510.16567/) (2025) - [Go Beyond Your Means: Unlearning with Per-Sample Gradient Orthogonalization](https://mlsanity.com/paper/2503.02312/) (2025) - [Mapping the Podcast Ecosystem with the Structured Podcast Research Corpus](https://mlsanity.com/paper/2411.07892/) (2024) - [Keyword-Guided Adaptation of Automatic Speech Recognition](https://mlsanity.com/paper/2406.02649/) (2024) - [Unsupervised Rhythm and Voice Conversion of Dysarthric to Healthy Speech for ASR](https://mlsanity.com/paper/2501.10256/) (2025) - [Do LLM Decoders Listen Fairly? Benchmarking How Language Model Priors Shape Bias in Speech Recognition](https://mlsanity.com/paper/2604.21276/) (2026) - [Sound Check: Auditing Audio Datasets](https://mlsanity.com/paper/2410.13114/) (2024) - [Tasks and Roles in Legal AI: Data Curation, Annotation, and Verification](https://mlsanity.com/paper/2504.01349/) (2025) - [Addressing Pitfalls in Auditing Practices of Automatic Speech Recognition Technologies: A Case Study of People with Aphasia](https://mlsanity.com/paper/2506.08846/) (2025) - [Model Science: getting serious about verification, explanation and control of AI systems](https://mlsanity.com/paper/2508.20040/) (2025) - [Whisper-CD: Accurate Long-Form Speech Recognition using Multi-Negative Contrastive Decoding](https://mlsanity.com/paper/2603.06193/) (2026) - [Transcription Policy as a Latent Variable: Activating Controllable Verbatim ASR with Word-Level Timing](https://mlsanity.com/paper/2607.18934/) (2026) - [ASR Benchmarking: Need for a More Representative Conversational Dataset](https://mlsanity.com/paper/2409.12042/) (2024) - [Quantization for OpenAI's Whisper Models: A Comparative Analysis](https://mlsanity.com/paper/2503.09905/) (2025) - [AsyncSwitch: Asynchronous Text-Speech Adaptation for Code-Switched ASR](https://mlsanity.com/paper/2506.14190/) (2025) - [SpeechIQ: Speech-Agentic Intelligence Quotient Across Cognitive Levels in Voice Understanding by Large Language Models](https://mlsanity.com/paper/2507.19361/) (2025) - [Conversational Speech Reveals Structural Robustness Failures in SpeechLLM Backbones](https://mlsanity.com/paper/2509.20321/) (2025) - [Beyond WER: Probing Whisper's Sub-token Decoder Across Diverse Language Resource Levels](https://mlsanity.com/paper/2509.25516/) (2025) - [From Dispersion to Attraction: Spectral Dynamics of Hallucination Across Whisper Model Scales](https://mlsanity.com/paper/2604.08591/) (2026) - [Identifying and typifying demographic unfairness in phoneme-level embeddings of self-supervised speech recognition models](https://mlsanity.com/paper/2604.22631/) (2026) - [The Case for Model Science: Verify, Explore, Steer, Refine](https://mlsanity.com/paper/2606.01189/) (2026) - [Said Aloud, Read Different: Cross-Modal Instability in Multimodal Models](https://mlsanity.com/paper/2608.27135/) (2026) - [CTC-DRO: Robust Optimization for Reducing Language Disparities in Speech Recognition](https://mlsanity.com/paper/2502.01777/) (2025) ## Abstract (arXiv) Speech-to-text services aim to transcribe input audio as accurately as possible. They increasingly play a role in everyday life, for example in personal voice assistants or in customer-company interactions. We evaluate Open AI's Whisper, a state-of-the-art automated speech recognition service outperforming industry competitors, as of 2023. While many of Whisper's transcriptions were highly accurate, we find that roughly 1% of audio transcriptions contained entire hallucinated phrases or sentences which did not exist in any form in the underlying audio. We thematically analyze the Whisper-hallucinated content, finding that 38% of hallucinations include explicit harms such as perpetuating violence, making up inaccurate associations, or implying false authority. We then study why hallucinations occur by observing the disparities in hallucination rates between speakers with aphasia (who have a lowered ability to express themselves using speech and voice) and a control group. We find that hallucinations disproportionately occur for individuals who speak with longer shares of non-vocal durations -- a common symptom of aphasia. We call on industry practitioners to ameliorate these language-model-based hallucinations in Whisper, and to raise awareness of potential biases amplified by hallucinations in downstream applications of speech-to-text models.