Mapbox, the mapping startup that competes with the likes of Google Maps and Here to power location services on apps like Snapchat and Foursquare, says that a version of its New York map calling the city ‘Jewtropolis’ was an act of human vandalism that has now been removed across the hundreds of apps that pull in Mapbox data and are used by over 400 million people.
“This is now 100 percent fixed and should have never happened. It’s disgusting,” Mapbox CEO and founder Eric Gundersen told TechCrunch. He says that Mapbox has built a large system that uses both AI and human checks to look out for vandalism. “We’re constantly scanning for this, and it’s an error on our part [to have missed it].”
The issue of the mis-named map using the anti-Semitic slur came to light earlier today, with a number of people flagging the glitch on social media and the issue also getting picked up by multiple media outlets.
Gundersen said that the company is still trying to figure out what specifically happened here (we’ll update as we learn more).
The company uses more than 130 sources of data to build its maps, and picks up incrementally more each time its maps are used in apps around the world (over 400 million people use its maps by way of third-party apps). While this is an interesting, and in many ways very clever, approach to building a data set as a smaller business that is miles away from the likes of a Google in terms of size and scale, this glitch highlights one of the pitfalls of relying on other people’s data.
Some are speculating that the vandalised map could have come from a vandalised version of OpenStreetMap, the open-source mapping project that is one of the data sources used by Mapbox. The reason: similarly defaced data was detected on OSM some weeks ago. But Gundersen cautioned against drawing a line between the two.
“The reality is that OpenStreetMap is just one part of what we use,” he said. “Mapbox is made from about 130 different sources of data.” He also compared Mapbox’s data set to Waze’s. “With over 400 million people using our maps, the more that they’re used, the more data we get.”
Mapbox pushed out a new updating system recently that allows it to update maps faster and to flag suspicious items. Gundersen said that the company is currently “digging into that and seeing if there is a flag off. You design for this never happening, and then it does.”
Mapbox has raised over $227 million in funding, with the most recent round being $164 million led by Softbank’s Vision Fund.
Update: Full statement from Mapbox reiterating what Gundersen said earlier:
“Mapbox has a zero tolerance policy against hate speech and any malicious edits to our maps. This morning, the label of “New York City” on our maps was vandalized. Within an hour, our team deleted and removed that information. The malicious edit was made by a source that attempted several other hateful edits. Our security team has confirmed no additional attempts were successful.
We build systems so this does not happen. Our maps are made from over 130 different sets of data, and we have a strong double validation monitoring system. Our preliminary root cause analysis shows that this act of hate speech was properly detected immediately and put into quarantine for human review.
Typically, our validation system prevents malicious edits from entering the system from any third party data source. Our AI system flags more than 70,000 map changes a day for human review. While our AI immediately flagged this, in the manual part of the review process a human error led to this incident.
Security experts are working to determine the exact origin of this malicious hate speech. We apologize to customers and users who were exposed to this disgusting attack.
We will continue to investigate this act and make appropriate changes to further limit the potential for future human error.”
Comment