Google to delete Street View source images at privacy watchdogs' request

European privacy watchdogs have demanded that Google delete the original images behind its Street View service. The company has said it will comply with the demand in the "long term".

Street View has raised privacy concerns wherever it has launched but the UK's privacy regulator the Information Commissioner's Office (ICO) has said that Google's privacy protections are good enough to protect people's privacy.

A committee of all 27 EU member states' privacy regulators, the Article 29 Working Party, has asked Google to ensure, though, that original images are destroyed once they have been used to create blurred images that the public can see.

When introducing Street View to Europe, Google undertook to automatically blur faces and vehicle number plates to protect the privacy of people photographed on the street.

It retains the unblurred originals, and says that these are vital for correcting errors. It has agreed, though, to delete those once they are no longer being used.

Article continues after advert

"The Article 29 Working Party has asked that we set a time limit on how long we keep the unblurred copies of panoramas from Street View, in a way that appropriately balances the use of this data for legitimate purposes with the need to deal with any potential concerns from individuals who might feature incidentally on the Street View imagery," said Peter Fleischer, Google's global privacy counsel, in a blog post.

Fleischer said that the original images were useful to Google because its automated software was not yet perfect.

"One of the technical challenges at stake with Street View – or any service that uses image detector software – is that the software sometimes makes mistakes, labelling part of the image as containing a face or a license plate when in fact it doesn't. While we like to think we've gotten pretty good at this stuff, we still have lots of these 'false positives'," he said. "We're constantly working on ways to improve our technology, and we are constantly training it to detect more of the relevant stuff, while reducing the number of 'false positives' it creates. To do this, though, we need access to the original unblurred copies of the images."