SYDNEY/NEW YORK, Oct 22 (Reuters) - Three years after

Meta shut down facial recognition software on Facebook

amid a groundswell of privacy and regulator pushback, the social

media giant said on Tuesday it is testing the service again as

part of a crackdown on "celeb bait" scams.

Meta said it will enroll about 50,000 public figures in a trial

which involves automatically comparing their Facebook profile

photos with images used in suspected scam advertisements. If the

images match and Meta believes the ad are scams, it will block

them.

The celebrities will be notified of their enrollment and can

opt out if they do not want to participate, the company said.

The company plans to roll out the trial globally from

December, excluding some large jurisdictions where it does not

have regulatory clearance such as Britain, the European Union,

South Korea and the U.S. states of Texas and Illinois, it added.

Monika Bickert, Meta's vice president of content policy,

said in a briefing with journalists that the company was

targeting public figures whose likenesses it had identified as

having been used in scam ads.

"The idea here is: roll out as much protection as we can for

them. 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," Bickert said.

The test shows a company trying to thread the needle of

using potentially invasive technology to address regulator

concerns about rising numbers of scams while minimising

complaints about its handling of user data, which have followed

social media companies for years.

When Meta shuttered its facial recognition system in 2021,

deleting the face scan data of one billion users, it cited

"growing societal concerns". In August this year, the company

was ordered to pay Texas $1.4 billion to settle a state lawsuit

accusing it of collecting biometric data illegally.

At the same time, Meta faces lawsuits accusing it of failing to

do enough to stop celeb bait scams, which use images of famous

people, often generated by artificial intelligence, to trick

users into giving money to non-existent investment schemes.

Under the new trial, the company said it will immediately

delete any face data generated by comparisons with suspected

advertisements regardless of whether it detected a scam.

The tool being tested was put through Meta's "robust privacy and

risk review process" internally, as well as discussed with

regulators, policymakers and privacy experts externally before

tests began, Bickert said.

Meta said it also plans to test using facial recognition

data to let non-celebrity users of Facebook and another one of

its platforms, Instagram, regain access to accounts that have

been compromised by a hacker or locked due to forgetting a

password.