Instagram is failing to enforce its own rules and allowing some of its most popular accounts to be attacked with abusive comments “with impunity,” according to a new report from the Center for Countering Digital Hate. The anti-hate group claims that Meta failed to remove 93 percent of the comments it reported to the company, including those containing racist slurs, threats of violence and other disturbing language that blatantly violate the social network’s rules.

CCDH researchers focused on five Republican and five Democratic lawmakers running for office this year. The group included Vice President Kamala Harris, Representative Nancy Pelosi, Senator Elizabeth Warren, Representative Marjorie Taylor-Greene, Senator Marsha Blackburn and Representative Lauren Boebert.

The researchers examined 1,000 comments that appeared on MPs’ Instagram posts between January and June this year and found that Meta took “no action” against the vast majority of these comments, 926 of which were still visible on the app a week after they were reported. The reported content included comments containing racist slurs and other racist language, calls for violence and other insults.

“We simulate the moment when someone reaches out and asks for help, and in fact Instagram’s failure to respond only exacerbates the damage that has been done,” CCDH CEO Imran Ahmed said in a briefing on the report.

The CCDH also found that many of the offensive comments came from “repeat offenders,” which Ahmed said created “a culture of impunity” on the platform. The report comes less than three months before the U.S. presidential election and notes that attacks on Harris, now running for president, appear to have “intensified” since she took the electoral slate. “Instagram failed to remove 97 of 105 offensive comments against Vice President Kamala Harris, representing a failure to respond to 92% of offensive comments against her,” the report said. It notes that Instagram failed to remove comments against Harris that used the N-word as well as gender-based slurs.

In a statement, Meta said it was reviewing the report. “We provide tools that allow anyone to control who can comment on their posts, automatically filter out offensive comments, phrases or emojis, and automatically hide comments from people who don’t follow them,” Meta’s head of women’s safety said in a statement. “We work with hundreds of safety partners around the world to continually improve our policies, tools, detection and enforcement, and we will review the CCDH report and take action against any content that violates our policies.”