10 billion "faces" from the net, the regulatory authorities deleted the AI face recognition

Government authorities ordered an AI face recognition venture to collect more than 10 billion face images from SNS and others.

Australian privacy officials have announced on March 3 that the US venture, Clear View AI, which provides AI face recognition services, has deleted the public's face image and related data.

Based on a huge amount of face images scraping (automatically collecting) from social media, etc., new regulations on the US venture "Clear View AI", which provide facial recognition services to investigative agencies and other events, etc.Measures.

However, the company should file an objection to the order, and it seems that there is still a song.

Regarding the recognition of the face, the day before the Australian authorities announced, it has just stated that Meta (Facebook) will delete more than 1 billion user's face image database.

Concerns about "large -scale monitoring" due to face recognition have become a global trend.

● Concerns that "being monitored"

Australian information commissioner and privacy commissioner Angeline Folk pointed out on March 3 in the announcement of the deletion order for AI venture in AI venture.

Folk is a privacy method in the country that prohibits a huge amount of face images from the Internet on the Internet by Clear View AI, which prohibits "collection of sub -information information without the consent of the person" and "collecting personal information by unreasonable means".He pointed out that he was violating.

He has also ordered the company to collect facial images from Australian citizens and stop creating data for face recognition, and destroy the data that has already been collected and created.

Folk further states:

This decision is based on the results of a joint survey conducted by the Australian Information Commissioner Secretariat (OAIC) and the ICO, the Information Commissioner Secretariat (ICO), a British privacy authorities.It is said that the British ICO conclusion has not yet come.

According to OAIC's certification, the service has been tested by several police, including the Australian Federal Police, from October 2019 to Clear View AI for this joint survey.It is said that it has been revealed.

OAIC is also investigating whether there was no problem in using the Federal Police Clear View AI.

The OAIC decision was dated on October 14, ordering the face image and data destruction within 90 days, and the report.

However, Clear View AI is complaining about the order.

According to Reuters and others, the company's lawyer argued that "Clear View AI has not been doing business in Australia and has no customers," and that the country's information commissioners have no jurisdiction.

● More than 10 billion databases

Clear View AI introduces the service on its own site and explains the face image database:

Clear View AI was established in 2017, but the New York Times reported in January 2020, gaining global attention.

According to the newspaper, the Clear View AI provides services to more than 600 law execution agencies, including the US Federal Bureau (FBI) and the Ministry of Land Security (DHS).

※参照：SNS投稿30億枚から顔データべース、警察に広がるAIアプリのディストピア（01/19/2020　新聞紙学的）

※参照：SNSから収集、30億枚の顔データベースが主張する「権利」（03/22/2021　新聞紙学的）

What we were concerned about was how to collect and scale the face image.

Clear View AI automatically collects facial images posted by users to social media such as Facebook, YouTube, Twitter, and Instagram, creating a database for face recognition.

Until now, the company has expressed the size of the collected facial image on its own site as "over 3 billion copies."However, in October 2021, the number was updated to more than three times the number of "over 10 billion copies".

According to a report to the U.S. Senate released by the US Accounting Inspection in 2016, the number of searches for the FBI face photo database was 411.9 million copies.

The size of the face image of the clear view AI is currently 24 times.

● Movement of a series of privacy authorities

In January 2020, reported by the New York Times, a group lawsuit against Clear View AI based on the Biological Information privacy Protection Law, which is the state law of Illinois, and in May of the same year, the US Freedom Union (ACLU) litigation.Woke up.

In addition, a series of suspension notifications due to platforms such as Facebook, Google, and Twitter, such as Facebook, Google, and Twitter, have been sent one after another.

Outside the United States, there are movements.

Hamburg's data protection and information disclosure commissioner (HMBBFDI) is a member of the German -based Hacker Group "Chaos Computer Club (CCC)" in January 2021, a researcher at Hamburg University, Matthew Marks.In response to the claim, the European Union (EU) was allowed to violate the Protection Law on Protection of the General Data (GDPR).

However, the commissioner has ordered the target to be deleted only to the data of the filtering of the petitioner, Marks.

In Sweden, the Privacy Protection Agency (IMY) used the Police Agency in February that the National Police Agency used the Clear View AI, and it would be illegal to the Criminal Data Law, and the agency was over 2.5 million chrones (approximately 33.4 million yen). is doing.

In the same month, clear view AI violates the Federal and Privacy Protection Law in Canada by the Federal Privacy Commissioner Secretariat, the Information Committee of Quebec, information and privacy commissioner secretariat in British Colombia and Alberta.Certified as having done it.

"Clear view is a large -scale monitoring and illegal," he said, providing services to customers in the country, stopping facial images, and removing face image data already collected.Was recommended.

On the other hand, Clear View AI stopped providing services to the police, but refused other recommendations.But in July, he notified the suspension of activities in the country.The company is still investigating the suspension and deletion of data.

In addition, in May human rights organizations such as "Privacy International" have violated the GDPR in May that in May, the Clear View AI has violated the GDPR against the United Kingdom, France, Italy, Greece and Austria.We have filed a claim calling for compiling measures.

The movement of privacy authorities so far has been a deletion order that Hamburg is limited to the data of the petitioner, the Sweden is not a clear view AI, but the Police Agency on the same side, a recommendation in Canada.It was limited.

In response, the order of the Australian authorities will be a full -fledged forced measure as a government agency.

● Meta deletes face recognition data for 1 billion people

On November 2, Meta (Facebook) Vice President AI, on November 2, explained on the official blog about the suspension of the company's face recognition system.

It automatically recognizes the face in the photo posted on Facebook, stops this service that makes it easier to tag name, and deletes data that is based on face recognition at once.

The reasons for this are rising concerns about face recognition and insufficient system development.

It can be said that the movement over the clear view AI that symbolizes concerns about face recognition and insufficient system development as Pesenti points out.

Already, IBM, Amazon, and Microsoft have already announced in June 2020 that their criticism of face recognition has grown in a series of withdrawal from services and delay to police.

※参照：IBM、Amazon、Microsoftが相次ぎ見合わせ、AIによる顔認識の何が問題なのか？（06/10/2020　新聞紙学的）

● Stopping as a society

Concerns about face recognition symbolized by Clear View AI are concerns about dystopia of large -scale real -time monitoring.

In the AI regulation bill released by the EU in April, real -time face recognition by law execution agencies such as police is in principle prohibited.

※参照：「すごく危ないAI」の禁止に潜む大きな「抜け穴」とは（04/25/2021　新聞紙学的）

In September, in September, the Yomiuri Shimbun reported that JR East Japan had used face recognition with security cameras for the source, and in the response, "Social consensus is still gained.On the same day, there was a turmoil that stopped using it on the same day.

The background of the clear view AI continuing to controversate in each country is such a concern about "large -scale monitoring of real -time", as well as a social media society assuming open information sharing, so to speak, "free rider (free riding (free) As a behavior, there is a collection of huge face images.

Is such a movement clearly stopped as a society?

The consequences of the Australian authorities are likely to be a clue to this issue.

（※2021年11月5日付「新聞紙学的」より加筆・修正のうえ転載）