Become a Member

Regulation TrackerLaws, orders and treaties read for their security duties and clocks Standards & FrameworksAttack taxonomies, control catalogues and frontier frameworks AI Incident DatabaseThe security incident record, with mechanism and lesson Research Paper LibraryThe papers that define the field, summarised OSS DirectoryRed teaming, guardrail, supply chain and evaluation tools Case StudiesProgrammes built, with measured outcomes AI Vulnerability TrackerCVEs and advisories for the components AI systems are built from

Latest in the tracker

New York Stop Addictive Feeds Exploitation (SAFE) for Kids Act and implementing rulesNew York's law restricting algorithmic feeds for minors takes effect on 25 January 2027 under final rules published in July 2026, with certified age assurance and $5,000-per-violation penalties.

Knowledge Base Regulation Tracker Standards & Frameworks OSS Directory AI Incident Database Research Paper Library Case Studies Society Articles News Jobs About Charter Become a Member

Home/AI Incident Database/OpenAI's Whisper fabricates text in transcripts, including invented medications and racial remarks, while used in hospitals

OpenAI's Whisper fabricates text in transcripts, including invented medications and racial remarks, while used in hospitals

Researchers found Whisper inventing sentences, including nonexistent drugs and violent content, in transcripts, even as hospitals adopted it for clinical notes without retaining the audio.

Severity: Highwhisperhallucinationmedical transcriptionnabla

At a glance

Organisation: OpenAI; Nabla and hospital systems
Industry: Healthcare
Country or region: United States
Occurred: 1 Jan 2024
Disclosed: 26 Oct 2024
Type of harm: Medical and safety-critical error, Misinformation and fabricated claims

All details 15

Kind of system: Speech and audio systems
Severity: High
Confirmation: Demonstrated by researchers
Model: Whisper
Model provider: OpenAI
Technique: Hallucinated phrases inserted during pauses and low-quality audio
Primary root cause: Model limitation or hallucination
Scale: Nabla's Whisper-based tool used for an estimated 7 million medical visits; researchers found hallucinations in about 1% of segments and 40% of those harmful
Outcome: Nabla said it deleted original audio for privacy, preventing verification; OpenAI warned against high-stakes use; the AP investigation drove FDA and hospital scrutiny of ambient scribes.

Sources

apnews.comhttps://apnews.com/article/ai-artificial-intelligence-health-business-90020cdf5fa16c79ca…

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.

Submit incident Report a correction or request removal

OpenAI's Whisper fabricates text in transcripts, including invented medications and racial remarks, while used in hospitals

Sources

Contribute

Related in AI Incident Database

Epic's sepsis prediction model misses most cases and floods clinicians with alerts