• AI facial recognition to be re-implemented on Meta platforms.
• 50k celebs to be opted-in by default.
• Global rollout from December as ‘security feature.’

In 2021, Meta shut down its AI facial recognition system amid privacy concerns. At the time the company pointed out that some features advantageous for blind and partially-sighted users of its platforms would lose some benefits, such as the automatic generation of image descriptions. The shuttering also closed off the option where users could be notified automatically if they appeared in images or video posted by others.

Meta paid a fine in Texas earlier this year amounting to $1.4 billion in settlement of a state-led lawsuit that accused the company of illegally collecting biometric data of its citizens prior to 2021, a hangover from the public outcry about the company’s practices.

However, in a bid to curtail the growing use of faked celebrity video in bogus advertising, Meta has announced that it is testing a reboot of its AI facial recognition system, aimed at around 50,000 public figures to take part, each of whom will be automatically enrolled, but given the option to opt-out of the trial.

The company currently faces legal action from public figures whose images appear without their permission or approval on Facebook and Instagram in posts created by third-parties. In many cases, these occurrences are so-called ‘celeb-bait,’ where famous faces are spoofed in AI-generated videos that promote goods or services unconnected with the well-known individual.

By making participation in the celebrities’ trial the default, and therefore requiring individuals to actively opt-out, Meta will have de facto approval in many cases, so will not be subject to claims from those affected that it is acting unlawfully.

AI facial recognition worldwide

A more widespread re-introduction of facial recognition is to take place globally from December in territories that are not subject to legislation designed to prevent companies from using the technology, so facial recognition features will not be permitted in some states in the US, the UK, and European countries that adhere to EU laws.

The company’s aim is to help its users recover lost or stolen account credentials. To do so, the platforms will have to store an uploaded video containing facial data of every user that wants to use the recovery feature. While the data thus gathered will likely not be used on the Meta platforms, its existence could be valuable to the company as a resource that it and its ‘chosen partners’ might want to make use of. In this context, partners are usually carefully chosen on the basis of whether or not they have the funds to purchase information.

In a briefing with journalists, Monika Bickert asserted the company’s stated aims in the first, celebrity focused part of the facial recognition reboot: “The idea here is: roll out as much protection as we can for [celebrities]. They can opt out of it if they want to, but we want to be able to make this protection available to them and easy for them.”

Meta has published no specific numbers of how many celebrities’ likenesses have been used in scams, although several high-profile figures like financial advisor Martin Lewis and film star Tom Hanks have had their images used in AI-generated advertisements that typically promote bogus investment schemes.

Speaking about the feature that will be released to the general public, Meta’s David Agranovich said on Monday, “[The] process is done in real time and is faster and much more accurate than manual human reviews.” Presumably the AI algorithms that will be used will not be subject to the same level of hallucinations (AKA errors) and false flags that bedevil the rest of the world’s AI platforms.

Meta reports that Facebook alone has 3.07 billion monthly active users. When the security facility based on facial recognition is released to non-celebrities from December, if even a small proportion of those who decide (or are allowed under local law) to upload a video selfie, the numbers partaking are likely to be high. Therefore, if a billion or two upload footage to verify their identity in the event of account hijack or forgotten credentials, Meta will be in possession of a great deal of personally identifiable information – quite literally – with which it will be able to protect its platform’s users, among other uses.

The cost of video storage and that of running AI facial recognition algorithms will be carried by Facebook as part of the free service it offers to the world’s citizens, thus cementing its place in global society as a figurehead of responsible personal data safeguarding and torchbearer for privacy.

Looking to learn more? Explore the Upcoming AI & Big Data Expo

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London.