The board previously said the policy “disproportionately restricts free expression” because while the term is “sometimes used by extremists to praise or glorify people who have died while committing violent terrorist acts,” there are also alternate meanings.

In a test, Meta said, removing the term when “paired with otherwise violating content” captured “the most potentially harmful content without disproportionality impacting voice.”

Correction: Meta said it’s implementing the Board’s recommendations, not seeking further guidance.