Draft:AI Incident Database
Submission declined on 1 November 2025 by Robert McClenon (talk).
Where to get help
How to improve a draft
You can also browse Wikipedia:Featured articles and Wikipedia:Good articles to find examples of Wikipedia's best writing on topics similar to your proposed article. Improving your odds of a speedy review To improve your odds of a faster review, tag your draft with relevant WikiProject tags using the button below. This will let reviewers know a new draft has been submitted in their area of interest. For instance, if you wrote about a female astronomer, you would want to add the Biography, Astronomy, and Women scientists tags. Editor resources
|  |
Comment: This draft does not have enough information to satisfy general notability. It may be resubmitted after expansion. Robert McClenon (talk) 18:11, 1 November 2025 (UTC)
| Owner | Responsible AI Collaborative |
|---|
The AI Incident Database (AIID) is an open, publicly searchable database that indexes real-world incidents and near-incidents involving artificial intelligence systems.[1][2][3] Launched publicly in November 2020 under the sponsorship of the Partnership on AI, it is maintained by the Responsible AI Collaborative, a non-profit organization chartered to advance the project.[4][5][6][7]
The database was modeled on incident reporting systems in aviation safety and computer security, notably the Common Vulnerabilities and Exposures (CVE) program, on the premise that a shared record of past failures helps practitioners avoid repeating them.[1][4] Its records are drawn largely from press reporting, with multiple articles about the same event ("reports") consolidated into a single incident entry.[8]
The database was introduced publicly in November 2020 as a project of the Partnership on AI.[4][3] Coverage at the time framed it as an attempt to create a permanent, citable record of AI failures analogous to the collections maintained in mature safety-critical industries.[3][2] An academic paper introducing the system was presented at the AAAI Conference on Artificial Intelligence in 2021.[1]
It is a widely referenced public source of AI incident data, and is used as an input by Stanford University's annual AI Index report and the MIT AI Risk Initiative's incident tracker.[9][10]
Scope and method
[edit]The database's editorial guidance scopes an incident as "an alleged harm or near harm event to people, property, or the environment where an AI system is implicated."[11] Entries are typically anchored to press reporting of a specific event, such as a wrongful arrest following a facial recognition error or the circulation of deepfake imagery.[5]
Since 2020, any member of the public may submit a candidate incident, which is reviewed by editors before publication.[5] Individual incidents are annotated with key basic information, such as the implicated deployers and developers of AI systems related to the incident, as well as using any of several contributed taxonomies covering, among other dimensions, the deployment sector, the risk domain, the type of harm, and the stage of the system lifecycle at which the failure arose.[12][13]
Because inclusion depends on an event having been reported and submitted, the collection is not an exhaustive record of AI harms but a compilation of those that became publicly newsworthy.[5]
Contents
[edit]The database grew from a small number of entries in its early years to more than 750 incidents and 3,500 supporting reports by September 2024, and to roughly 1,460 incidents by 2026.[14]
Analyses of the collection have found a sustained rise in recorded incidents, with roughly half of the entries indexed as of late 2024 dating from 2022 or later.[5] Stanford's AI Index reported 233 documented incidents for 2024 and 362 for 2025.[9] The growth has been uneven across categories: incidents involving autonomous vehicles were more prominent in earlier years, while fraud, scams and disinformation involving generative AI account for much of the more recent increase.[5][8]
Limitations
[edit]Commentators and researchers have identified constraints inherent to the database's method. Because entries derive from public reporting and voluntary submission, the dataset is subject to sampling bias and reflects media attention as well as underlying incident rates.[10][5] An analysis by the Center for Security and Emerging Technology found that the AI Incident Database and comparable repositories record mostly high-level information (such as the names of the systems and deployers involved) and rarely capture granular technical detail about model properties, training data or evaluation, in part because such information is seldom published by developers.[15] Researchers examining incident data for agentic systems have made a similar point, noting that publicly sourced databases cannot supply activity logs or system internals.[16]
See also
[edit]References
[edit]- 1 2 3 McGregor, Sean (2021-05-18). "Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database". Proceedings of the AAAI Conference on Artificial Intelligence. 35 (17): 15458–15463. doi:10.1609/aaai.v35i17.17817. ISSN 2374-3468.
- 1 2 Simonite, Tom. "Don't End Up on This Artificial Intelligence Hall of Shame". Wired. ISSN 1059-1028. Retrieved 2026-07-30.
- 1 2 3 Ferdowsi, Samir (2020-11-23). "This Database Is Finally Holding AI Accountable". VICE. Retrieved 2026-07-30.
- 1 2 3 McGregor, Sean (2020-11-18). "When AI Systems Fail: Introducing the AI Incident Database". partnershiponai.org. Partnership on AI. Retrieved 2026-07-30.
Avoiding repeated AI failures requires making past failures known. Therefore, today we introduce a systematized collection of incidents where intelligent systems have caused safety, fairness, or other real-world problems: The AI Incident Database (AIID).
- 1 2 3 4 5 6 7 Peek, Katie (2026-01-16). "What experts can learn by tracking AI harms". Bulletin of the Atomic Scientists. Retrieved 2026-07-30.
- ↑ "The details about who and why for database". AI Incident Database. Retrieved 2026-07-30.
- ↑ "About". Responsible AI Collaborative. Retrieved 2026-07-30.
- 1 2 Booth, Harry (2026-01-16). "What the Numbers Show About AI's Harms". Time. Retrieved 2026-07-30.
- 1 2 Stanford Institute for Human-Centered Artificial Intelligence (2026). Artificial Intelligence Index Report 2026 (Report). Retrieved 2026-07-30.
- 1 2 "AI Incident Tracker". MIT AI Risk Initiative. Retrieved 2026-07-30.
- ↑ "Editor's Guide". incidentdatabase.ai. Retrieved 2026-07-30.
- ↑ "List of taxonomies". incidentdatabase.ai. Retrieved 2026-07-30.
- ↑ "Adding Structure to AI Harm". Center for Security and Emerging Technology. Retrieved 2026-07-30.
- ↑ Mengesha, Isaak; Owen, Branwen; Collins, Charlie; Wong, Tina; Mylius, Simon; Slattery, Peter; McGregor, Sean (2026). "A pragmatic classification framework for AI incident monitoring". arXiv:2604.21412 [cs.CY].
- ↑ Hoffmann, Mina; Frase, Heather (January 2025). AI Incidents: Key Components for a Mandatory Reporting Regime (PDF) (Report). Center for Security and Emerging Technology. Retrieved 2026-07-30.
- ↑ Ezell, Carson; Roberts-Gaal, Xavier; Chan, Alan (2025). "Incident Analysis for AI Agents". arXiv:2508.14231 [cs.CY].
- provide significant coverage: discuss the subject in detail, not just brief mentions or routine announcements;
- are reliable: from reputable outlets with editorial oversight;
- are independent: not connected to the subject, such as interviews, press releases, the subject's own website, or sponsored content.
Please add references that meet all three of these criteria. If none exist, the subject is not yet suitable for Wikipedia.