keep scrolling down

Artificial Intelligence technologies typically utilize machine learning algorithms, neural networks, and large datasets to perform tasks such as data analysis, pattern recognition, and decision-making with minimal human intervention.

1/5

The new EU AI Act classifies each AI use into one of three risk levels: unacceptable, high, or low.

The bigger the dot , the higher the number of real-world incidents associated with that use.

2/5

Unacceptable uses are strictly forbidden.

High-risk uses , crucial in domains such as safety and education, walk a fine line between importance and risk.

3/5

Then there are the low-risk uses , those seemingly benign chatbots and innocuous recommender systems. They appear benign but harbor potential dangers.

4/5

Click icon Discover the hidden dangers of low-risk uses in all tech by clicking on . Even the most innocuous applications can unveil unexpected and deleterious consequences.

5/5


Methodology

This visualization showcases a dataset of 380 real-world uses of AI, linked to incidents reported in the news and sourced from the AI Incident Database. As of March 2024, there were 649 incidents reported from 3,412 news articles, primarily from the US. Most of these incidents were reported between 2013 and 2024. To create the dataset, we initially downloaded all 649 incidents. After removing duplicate entries, we were left with 639 distinct incidents. We analyzed ther key variables such as title, description, AI deployer, AI developer, and harmed subjects. We then paraphrased the incidents into 380 AI uses, and consulted experts to list their risks and benefits.



Research articles

The Atlas of AI Risks: Enhancing Public Understanding of AI Risks, AAAI HCOMP

The Atlas of AI Incidents in Mobile Computing: Visualizing the Risks and Benefits of AI Gone Mobiles, ACM Mobile HCI

Good Intentions, Risky Inventions: A Method for Assessing the Risks and Benefits of AI in Mobile and Wearable Uses, ACM Mobile HCI

Decoding Real-World AI Incidents, IEEE Computer



Team

Data Analysis

Sanja Šćepanović, Julia De Miguel Velazquez, Marios Constantinides, Daniele Quercia, Andrés Gvirtz


Data Visualization

Edyta Bogucka, Sanja Šćepanović, Daniele Quercia