Five companies appear in more than a third of all artificial intelligence incidents logged in the AI Incident Database, according to a Surfshark analysis of records dating to 2010. The database collects, verifies and classifies user reported events in which an AI system caused or nearly caused harm to people, property or the environment.

Facebook was named in 48 incidents, the highest count in the dataset. Tesla followed with 40, OpenAI with 39, Google with 33 and Amazon with 24. Events involving those five companies made up more than 35 percent of every incident recorded as of the collection date of June 16, 2023.

The pattern behind each company's total differs. Facebook logged no incidents before 2014, after which its yearly count climbed to a peak of 13 in 2020, giving it about 9 percent of all entries. Tesla registered its first incident in 2016 and has averaged roughly seven per year since, accounting for close to 8 percent of the database. OpenAI recorded a single incident in 2016 and none again until 2019, followed by a sharp climb through 2023 that placed it slightly below 8 percent of all entries and made it the most frequently named of the five companies that year.

Entries reflect alleged harms drawn from media coverage and user submissions rather than adjudicated findings. Examples cited in the analysis include a claim that Facebook's advertising algorithm displayed blue collar job listings unevenly by age and gender, and video that reportedly showed a Tesla running FSD Beta 11.4.1 failing to yield to a pedestrian at a marked crosswalk.

Source: Surfshark - https://surfshark.com/research/chart/ai-incidents-of-companies