In leaked memo to advertisers, Facebook's top ad exec promises an external audit of its safety tools and practices and tries to talk fleeing advertisers off the ledge

In a new memo sent on Friday, Facebook's VP of global marketing solutions Carolyn Everson assured advertisers that the company is serious about addressing the major concerns of civil-rights groups leading a boycott of the platform to protest hate speech and divisive content.
Everson said that the company was committing to a third-party audit of its Community Standard Enforcement Report (CSER). She also said that Facebook was considering nine product recommendations by the boycott organizers.
Everson said that Facebook is exploring how to extend the offerings in its Brand Safety Hub more broadly, and was also exploring giving moderators better tools for moderating content and membership to Groups.
Visit Business Insider's homepage for more stories.

Facebook's top ad executive assured advertisers that the company is serious about addressing the major concerns of civil-rights groups leading a boycott of the platform to protest hate speech and divisive content.

In a new memo sent on Friday, Facebook's VP of global marketing solutions Carolyn Everson said that the company was committing to a third-party audit of its Community Standard Enforcement Report (CSER). She also said that Facebook was considering nine product recommendations by the boycott organizers, and outlined the steps it had already taken to that end.

A Facebook spokeswoman confirmed the authenticity of Everson's memo but did not provide additional comment.

Everson said that while the report already includes figures on the total amount of hate speech that the company removes, and the proportion of hate speech that it detects proactively before it is reported, it would now also add the prevalence of hate speech to the report over the coming year. She added that Facebook was also working with the industry group Global Alliance for Responsible Media (GARM) to identify appropriate brand safety audit requirements, and would then work with the Media Ratings Council on an audit of its brand safety tools and practices.

Facebook CEO Mark Zuckerberg and his deputies, including Everson, have been making the rounds to assure top advertisers that the company is making progress on stamping down hate speech. The social media giant is reaching out to ad agencies as it faces mounting pressure over its content moderation policies and its handling of President Donald Trump's posts in the aftermath of George Floyd's death in police custody.

In the memo, Everson also said that Facebook is exploring how to extend the offerings in its Brand Safety Hub more broadly, after launching several new tools last month to give advertisers more control over where their ads appear within in-stream placements. She did note, however, that Facebook was unable to provide complete access to the content it deletes, but did provide metadata like the publisher name, video title, and ad impressions.

"We have absolutely no desire to profit from hate or misinformation of any kind, and while there are substantial technical challenges to extending the offerings in the Brand Safety Hub more broadly, we are exploring what is possible," she wrote.

Facebook is also exploring giving moderators better tools for moderating content and membership to Groups on its platform, and ways to make moderators more accountable for the content in groups they moderate, according to the memo.

Everson also pointed to the new policies that Facebook announced on June 26 to address growing calls from civil rights groups to clamp down on racist and hurtful content on its platform. In a livestreamed address, Zuckerberg said that the company would expand its ad policy to prohibit claims that people with a specific race, ethnicity, sexual orientation or religion are a threat to anyone else. Facebook said it would also begin labeling content deemed "newsworthy" like a speech from a politician — even if it violates the company's policies.

Read Everson's memo below:

Hi All,

I wanted to update you on several new developments and provide you with a response to the 9 product recommendations by the boycott organizers. We have significant work underway to address the major concerns expressed by those calling for the boycott but acknowledge there is much more work to do.

Mark just announced some important changes:

Changing our policies to better protect immigrants, migrants, refugees and asylum seekers. We've already prohibited dehumanizing and violent speech targeted at these groups - now we're banning ads suggesting these groups are inferior or express contempt, dismissal or disgust directed at them.

Banning posts that make false claims about ICE agents checking for immigration papers at polling places, or other threats meant to discourage voting. We will use our Election Operations Center to work with state election authorities to remove false claims about polling conditions in the 72 hours leading up to election day.

Labeling content that we leave up because it is deemed newsworthy, so people can know when this is the case. We'll allow people to share this content to condemn it, just like we do with other problematic content, because this is an important part of how we discuss what's acceptable in our society -- but we'll add a prompt to tell people that the content they're sharing may violate our policies.

These changes are the result of direct feedback from the civil rights community collected through our civil rights audit. That audit has been led by Laura Murphy, a noted civil rights expert, Megan Cacace, a partner at the respected civil rights law firm of Relman & Colfax.

In addition to the changes that Mark announced, I want you to know that we are also committing to the following important steps:

A third-party audit of our Community Standard Enforcement Report (CSER) which will be completed by a Big 4 Firm, and will include the incidence of violating content.
Adding the prevalence of hate speech to CSER over the coming year (CSER already includes the total amount of hate speech that we remove, as well as the proportion of hate speech that we detect proactively before it is reported).
We are working with the Global Alliance for Responsible Media (GARM) to identify appropriate brand safety audit requirements. We are committed to then working with the Media Ratings Council on an audit of our brand safety tools and practices.

I also want to respond to the 9 product recommendations from the boycott organizers to describe work that is already underway and areas where we are exploring further changes.

The first set of recommendations ask that we provide more support to people who are targets of racism, antisemitism, and hate.

The first recommendation is to create a separate moderation pipeline staffed by experts on identity-based hate for users who express they have been targeted because of specific identity characteristics. Today, hate speech reports on Facebook are already automatically funneled to a set of reviewers with specific training in our identity-based hate policies in 50 markets covering 30 languages. In addition, we consult with experts on identity-based hate in developing and evolving the policies that these trained reviewers enforce.

The second recommendation is to put targets of hate and harassment in touch with Facebook. We agree it is important to provide support to people who have been targeted on our platform. Our approach, developed in consultation with experts, is to follow up with people who report hate speech and tell them about the actions we've taken. We also provide user controls that allow people to moderate comments on their posts, block other users, and control the visibility of their posts by creating a restricted list. We are open to exploring ways to put people who report hate speech in touch with additional resources to find support, as we do do when people report content for suicide and self-injury. That could include helping people connect with organizations who counter hate speech and harassment in their communities.

The third recommendation is to provide more granular information about identity-based hate speech reports. We are committed to continuing to improve transparency about our Community Standards enforcement, and as I mentioned above, we intend to include the prevalence of hate speech in future CSER reports, pending no further complications from COVID-19.

We also hear the request to understand the kinds of hate speech that people report to us - including whether a report is based on a specific protected characteristic like ethnicity. As we have been improving our ability to recognize hate speech proactively across languages and geographies, we have prioritized speed and our ability to remove hate speech. While that has helped us tackle hate in the aggregate, we understand there are gaps felt for how hate can impact individuals or specific groups. We deeply value the views and experiences of groups wounded by hate. We will continue to support efforts to understand how particular communities are targeted by hate on our platform and the connections between online and offline hate, but we don't yet have specific detail on the frequency of attacks on specific groups on our platform.

The next set of recommendations focuses on how we can ensure we do not generate revenue from hate speech or misinformation - including preventing ads from appearing near content labeled as hate or misinformation, telling advertisers how often their ads have appeared next to this content, providing refunds to those advertisers, and providing an audited transparency report.

In response, we first want to emphasize again that hate speech violates our community standards and we remove it as soon as possible. We also work with third-party fact checkers to identify misinformation, and we put prominent labels and reduce the distribution of content or disapprove related ads if the content is rated false or partly false on our platform.

We already take some steps to help advertisers understand how their ads show up on our services, and to provide refunds in certain circumstances. For example, we have built a Brand Safety Hub where advertisers can review publishers, individual in-stream videos, and instant articles in which their ads were embedded. We refund advertisers when ads run in videos or instant articles that are later rated false or partly false by our third party fact checkers. We are not able to provide complete access to content that has been deleted, including for violating our Community Standards, but we do provide metadata such as the publisher name, video title, and ad impressions in these circumstances.

We have absolutely no desire to profit from hate or misinformation of any kind, and while there are substantial technical challenges to extending the offerings in the Brand Safety Hub more broadly, we are exploring what is possible. It will take some time for us to do this, and to obtain more detailed feedback about the systems and controls that would work for our advertisers and our community.

With respect to transparency and audits, our Community Standards Enforcement Reports provide extensive information about our efforts to keep our community safe. We've also set the standard in our industry by publishing regular transparency reports so people can hold us accountable for progress. We will continue investing in this work and will commit whatever resources are necessary to improve our enforcement.

We also have received certification from independent groups, like the Digital Trading Standards Group which specifically examines our advertising processes against JICWEBS' Good Practice Principles and is a requirement for achieving the Interactive Advertising Bureau's Gold Standard. We will continue to work with advertising industry bodies such as the Global Alliance for Responsible Media and the Media Ratings Council to audit our brand safety tools and practices.

The last set of recommendations relate to improving the safety of private Groups on Facebook. Our team of 35,000 safety and security professionals, including 15,000 content reviewers, actively review potentially violating content today, including in private Groups. Our proactive, AI-based detection tools are also used to identify hateful content and Groups that aren't reported to us. We clearly have more work to do here.

We believe that Group moderators already play a vital role in maintaining their Groups, and we therefore hold them to a higher standard. If moderators post or permit posting of violating content, the Group incurs penalties that can result in the Group being removed from Facebook (or, for less severe violations, prevented from being recommended to others). We are exploring providing moderators with even better tools for moderating content and membership, as well as ways to make moderators more accountable for the content in groups they moderate, such as providing more education on our Community Standards and increasing the requirements on moderatoring potential bad actors.

In summary, we have thoroughly reviewed the 9 product recommendations from the boycott organizers and know we share the goal of limiting hate and divisive content on our platform.

To reiterate, our immediate next steps include:
A third-party audit of our Community Standard Enforcement Report (CSER) which will be completed by a Big 4 Firm, and will include the incidence of violating content.
Adding the prevalence of hate speech to CSER over the coming year (CSER already includes the total amount of hate speech that we remove, as well as the proportion of hate speech that we detect proactively before it is reported).
We are working with the Global Alliance for Responsible Media (GARM) to identify appropriate brand safety audit requirements. We are committed to then working with the Media Ratings Council on an audit of our brand safety tools and practices.
Last, as Mark announced today, we are updating our policies effective immediately:
Changing our policies to better protect immigrants, migrants, refugees and asylum seekers. We've already prohibited dehumanizing and violent speech targeted at these groups - now we're banning ads suggesting these groups are inferior or express contempt, dismissal or disgust directed at them.
Banning posts that make false claims about ICE agents checking for immigration papers at polling places, or other threats meant to discourage voting. We will use our Election Operations Center to work with state election authorities to remove false claims about polling conditions in the 72 hours leading up to election day.
Labeling content that we leave up because it is deemed newsworthy, so people can know when this is the case. We'll allow people to share this content to condemn it, just like we do with other problematic content, because this is an important part of how we discuss what's acceptable in our society -- but we'll add a prompt to tell people that the content they're sharing may violate our policies.

This isn't work that ever finishes. Hate is an insidious feature of every society, and that is reflected across all platforms. But we also believe in our responsibility to help change the trajectory of hate speech - and while we know we can't eradicate it, we will continue to do everything in our power to shatter its presence on our platform. We are proud of our efforts to improve proactive detection of hate speech, and we'll continue to invest there. We won't permit hate to have a home on Facebook, and we want to work with you as we have through the years, to continue improving on our efforts. We will continue to be your partners, and we will continue to value the feedback you give us on where you see us falling short.

Thanks so very much
Carolyn