Tech giant Meta is facing allegations that it has misled Australians by exaggerating its response to disinformation and misinformation on Facebook.

The ABC can reveal the $US1.3 trillion ($AU1.9 trillion) company is the subject of a formal complaint, which alleges Meta is not living up to claims contained within its most recent transparency report.

The seven-page submission by online safety research group Reset.Tech, states small changes such as rewording can circumvent Facebook’s automated labelling of known falsehoods.

The complaint outlines the treatment of a post with the message: “Did you know ONLY 2 countries are still considered Sovereign? Russia & Australia.”

The statement was found to be false by Meta’s fact-checkers and labelled with a warning, along with any identical or “near-identical” posts Meta detected using artificial intelligence.

However, problems emerged when the message was reworded slightly to say, “There’s only two countries that are still sovereign in the world. Australia and Russia.”

The second post — and others like it — were left unlabelled, even though it contains the same falsehood.

Labelled by Meta: Russia and Australia are the only two countries still considered sovereign.

Not labelled by Meta: There’s only two countries that are still sovereign in the world. Australia and Russia.

Misinformation loopholes

Under Australia’s voluntary code of practice for online platforms dealing with misinformation, companies are asked to publish annual transparency reports to show how they are dealing with viral falsehoods.

Reset.Tech says it is concerned Meta is glossing over the labelling issue in its transparency report.

Meta’s report states that the company “applies a warning label to content found to be false by third-party fact checking organisations”, adding that nine million pieces of Facebook content had been labelled in Australia throughout 2022.

While nine million posts may sound impressive, according to Reset.Tech it is only a fraction of the whole, and the majority of known falsehoods could well be going unlabelled on Facebook.

“In actuality [they’re] only labelling a tiny sub-portion of it … there’s a real gap here between statement and practice,” executive director of Reset.Tech, Alice Dawkins said.

“I don’t think a company in any other sector could get away in the public transparency report, saying one thing and then actually doing another, much narrower, modest thing.”

Complaint considered by industry body

Meta denied the claims in its transparency report were misleading.

In email correspondence with Reset.Tech, obtained by the ABC, a Meta spokesperson said:

“Meta does not assert, nor does the statement convey, that Meta detects and labels ‘all variations of fact-checked falsehoods’. 

“Rather, the … language is clear that Meta proactively identifies content that ‘matches’ or is ‘near-identical’ to content that has been rated as ‘false’ by a fact-checking partner.

“We are careful in applying fact-checker ratings only to near-identical versions of content they have already rated because there may be meaningful semantic differences in content making similar claims.”

According to Meta, Reset.Tech has misinterpreted the language in its transparency report to be claiming more than it is.

Crucially, there seems to be no dispute between the two parties about which falsehoods attract and escape labelling.

The complaint against Meta was submitted on Tuesday to the Digital Industry Group Inc (DIGI), which represents several major tech platforms in Australia, including Meta.