The US-based social media behemoth, Facebook “misplaced” guidance on an important exemption to its rules on dangerous individuals and organizations for three years, the company’s independent oversight board revealed.
The board, which was formed by Facebook to rule on a small slice of contentious content decisions, said it had overturned the company’s original removal of an Instagram post encouraging people to talk about the solitary confinement of Mr. Abdullah Ocalan, a founding member of the Kurdistan Workers’ Party (PKK).
The company said the content should never have been removed, but it also said that after it selected the case, Facebook found a relevant piece of its internal rules had “inadvertently not transferred” to a new review system in 2018.
The guidance made an exception to Facebook’s rules, which bans support or praise of individuals or organizations it designates as dangerous, to enable discussion on the conditions of confinement.
The social media giant has long been under scrutiny over what is allowed on its platforms and has been criticized by the board for a lack of transparency around its rules. The board said it was “concerned” that Facebook had lost an important policy exemption for this time and it could have led to other posts being wrongly taken down.
The board said the guidance, which was not shared with Facebook’s policy team, was developed in 2017 partly in response to concerns about the conditions of Ocalan’s imprisonment.
A company spokeswoman declined to respond to questions about how the policy was lost. The board said Facebook was conducting a review of how it failed to transfer the guidance but said it was not “technically feasible” to determine how many pieces of content were taken down while the guidance was not available.
Facebook had restored the content before the board’s decision. The board has recommended the social media giant to publish the results of its review, including descriptions of any other lost policies.
Related: Google removed 9 malicious apps from Play Store over data theft