AI Incidents

The widespread deployment of Artificial Intelligence solutions increases the vulnerability of business and social systems to AI errors. There is comprehensive consolidation of AI induced incidents at the AI Incidents Database which helps to understand the rate, consequences and root causes of AI errors.

Report incidents through to . The following chart includes an estimate of the as yet unreported incidents using the average reporting lag (between incident and updating and publishing) the AI Incidents Database of 67 days. The estimated total number of incidents at a date is the number of incidents already reported at that date divided by the proportion of the waiting time distribution that has passed from that date to the database publication date.

AI Incident Trend

Major AI deployers

Cumulative incidents by the major AI deployers, as expected the major technology companies are associated with the majority of incidents. In assigning incidents, the deployer is accountable for the misuse of AI technologies by their users. In the following summarisation:

• Google includes Youtube;
• Microsoft includes office as well as technology solutions such as Azure, Copilot and Github;
• OpenAI includes ChatGTP;
• xAI includes Twitter;
• Amazon includes Amazon Web Services (AWS);
• Russian Government includes suspected russian state actors;
• US Government includes all federal departments;
• US Republican Party includes the Trump campaign.

Incidents for each deployer can be investigated at the AI Incident Database.

AI Incidents Database Reporting Lag

The shape of the waiting time distribution is estimated from the median of the observed reporting lag from the AI Incidents Database using all incidents from 2022 to 2024, which excludes earlier years where incident reporting practices may have been slower, and recent periods where sufficient time has not passed for all incidents to be reported.