Meta apologized Thursday, saying it had fixed a "bug" that caused some Instagram users to report excessive amounts of violent and graphic content being recommended on their personal "Reels" pages. "We have fixed a bug that resulted in some users seeing content in their Instagram Reelsfeed that should not have been recommended," a Meta spokesperson said in a statement. "We apologize for this error."
The statement comes after some Instagram users expressed concerns on various social media platforms about the recent spate of violent and "not safe for work" content recommendations.
Some users claimed that they saw such content even though Instagram's Sensitive Content Controls were enabled to the highest moderation setting.
In accordance with the media policy, the company works to protect users from disturbing images and to remove particularly violent or graphic content.
Banned content may include "videos depicting dismemberment, visible entrails, or charred corpses," as well as "sadistic comments about images depicting human and animal suffering."
However, Meta said it would also allow graphic content if it helps users condemn and raise awareness of important issues such as human rights violations, armed conflict or acts of terrorism.
On Wednesday night, CNBC saw several posts on Instagram in the United States that appeared to include dead bodies, images of injuries and violent attacks. These posts are marked as "sensitive content."
According to Meta's website, it uses in-house technology and a team of more than 15,000 moderators to help detect disturbing images.
The technology includes artificial intelligence and machine learning tools that help prioritize posts and remove "the vast majority of violating content" before users report it, the website said.
In addition, Meta strives to avoid recommending content on its platform that may be "low quality, offensive, sensitive or inappropriate for young audiences."
However, InstagramReels' mistake comes after Meta announced plans to update its moderation policy to better promote free expression.
In a statement released on January 7, the company said it would change the way it enforces some content rules to reduce errors that lead to users being censored.
Meta said this includes shifting its automated systems from scanning "all policy violations" to focusing on "illegal and serious policy violations such as terrorism, child sexual exploitation, drugs, fraud and scams." For less serious policy violations, the company added that it will rely on users to report issues before taking action.
Chris Kelly, former chief privacy officer of Facebook, said Meta is returning to its tradition of free speech. Meta, meanwhile, said its systems had downgraded too much content based on predictions that it was "likely" to violate standards, and that it was "eliminating most of the downgrades."
CEO Mark Zuckerberg also announced that the company would allow more political content and change its third-party fact-checking program to a "community notes" model, similar to the system on Elon Musk's X platform.
The moves are widely viewed as Zuckerberg's efforts to improve relations with U.S. President Donald Trump, who has criticized Meta's moderation policies in the past.
According to a spokesperson for Meta on X, the CEO visited the White House earlier this month “to discuss how Meta can help the government defend and advance U.S. technology leadership abroad.”
Meta is laying off 21,000 employees, nearly a quarter of its workforce, as part of a wave of tech layoffs in 2022 and 2023, with much of its Civic Integrity and Trust & Safety teams affected.