Tech giants using AI to remove billions of fake accounts

Google and Meta, the parent company of Facebook, have used artificial intelligence to remove billions of fake accounts and ads in a bid to crack down on internet fraud, an Oireachtas committee has heard.

Both companies detailed multibillion-dollar investment in system security to the Oireachtas finance committee which is examining how authorised push payment (APP) fraud is affecting Irish consumers.

The committee has described APP as the use of sophisticated techniques to trick victims into making real-time payments or to transfer funds.

A lucrative business for crime gangs, it concerns the use of fake social media and other online ads that coax users into paying for a product or service that does not exist.

Student accommodation: A market plagued by bed shortages, rising rents and 51-week leases

---

An attempt to buy the next election with a giveaway budget would be a gross miscalculation

---

The outlook for interest rate cuts is changing - so should Irish borrowers be worried?

---

IDA gives Intel €30m in aid package

---

Addressing the committee on Tuesday, Google’s government affairs and public policy manager in Ireland, Ryan Meade, said the tech platform used a mixture of technology, including sophisticated machine learning and human review to enforce its policies.

AI-powered “classifiers” flag potentially harmful content and globally last year Google removed 5.2 billion ads and restricted another 4.3 billion.

“Most of these actions took place before the ad was seen by a user. Content moderation at this scale is only possible with AI,” Mr Meade told the committee.

It also suspended more than 6.7 million advertiser accounts and blocked or restricted ads from serving on over 1.57 billion publisher pages across more than 143,000 publisher sites, up from 63,000 in 2021.

In the first half of this year, Meta took down 1.1 billion fake accounts. Dualta Ó Broin, its head of public policy for Ireland, said that by its nature fraud is difficult to spot and perpetrators are always looking out for ways to subvert rules and safeguards.

“Just as it is unlikely that fraud will ever be eradicated in society at large, it is unlikely we will ever be able to completely eradicate it online,” he said.

Since 2016 Meta has spent $20 billion (€18.8 billion) on safety and security, one quarter of which was spent last year alone. As well as human reviews, it uses AI to detect suspicious activity, focusing on behaviour rather than content and assessing whether page creations are made by humans and not bots.

Tuesday’s hearing was also addressed by the online retail giant Amazon which pointed out that while scams were not typically perpetrated across its services, fraudulent impersonation of the company was a problem.

“We have zero tolerance for scammers who attempt to impersonate Amazon,” said its head of public policy in Ireland, Ed Brophy.

Last year it invested more than €1.2 billion and employed more than 15,000 people in combatting fraud and other forms of abuse.

It also “initiated takedowns” of more than 20,000 phishing websites and 10,000 phone numbers being used for impersonation scams, as well as reporting hundreds of bad actors across the world to law enforcement.

However, Sinn Fein finance spokesman Pearse Doherty challenged the companies on the issue of fraudulent ads slipping through security nets, noting in particular a Google search that resulted in a fake, misspelt Dunnes Stores ad in a premium search position.

He said conspicuously fraudulent ads appearing in print newspapers would not be tolerated “and neither should it be for social media companies”.

During a robust exchange Mr Doherty asked the executives for their views on UK proposals that in cases where online companies have benefited financially from fraudulent advertisements, they could be held partially accountable for the fraud.

Philip Milton, a member of Meta’s UK public policy team, said he was not aware of the proposals but that their money was “best spent on trying to prevent this stuff from happening”.

“Yes things get through our protections and they always will because this is an adversarial space and these are highly organised criminal gangs,” he said.

Asked by Senator Aidan Davitt for specific data on reporting activity to An Garda Síochána, the three tech companies said they only had global numbers available.

Both Meta and Google said while they deal with protecting consumers against online content, they may not actually be privy to a related fraud that takes place off their sites.

Amazon said specific information about their interaction with Irish regulators would be contained in a forthcoming transparency report.