Exclusive: Twitter executive says moving fast on moderation as harmful content rises

Dec 2 (Reuters) – Elon Musk’s Twitter is leaning heavily on automation to moderate content, doing away with some manual reviews and favoring restrictions on distribution rather than removing some speech outright, raising its trust and security told Reuters.

Twitter vice president of trust and safety product Ella Irwin said Twitter is more aggressively restricting abuse-prone hashtags and search results in areas including child abuse, regardless of the potential effects on the “benign use” of those words.

“The biggest thing that has changed is that the team is fully empowered to move fast and be as aggressive as possible,” Irvine said on Thursday.

His comments came after Musk announced an amnesty for accounts suspended under the company’s previous leadership that had not broken the law or engaged in “serious spam” as researchers found hate speech on the social media service. reporting an increase.

company faced tough questions about his ability and willingness to moderate harmful and illegal content as Musk cut half of Twitter’s staff and issued an ultimatum to work longer hours that resulted in the loss of hundreds more employees.

and advertisersTwitter’s main revenue source has fled the platform over concerns about brand safety.

on Friday, musk swore “Important reinforcement of content moderation and protection of freedom of speech” in a meeting with French President Emmanuel Macron.

Irwin said that Musk encouraged the team to worry less about how their actions would affect user growth or revenue, adding that security was the company’s top priority. “She insists that every day, multiple times a day,” she said.

According to former employees familiar with that work, the security approach described by Irwin at least partially reflects an acceleration of changes Twitter was already planning since last year to deal with hateful conduct and other policy violations.

One approach rooted in the industry’s mantra of “freedom of expression, not freedom of access” involves omitting certain Tweets that violate company policies, but preventing them from appearing in places like the home timeline and search.

Twitter has long deployed such “visibility filtering” tools around misinformation and had already incorporated them into its official hateful conduct policy prior to the Musk acquisition. The approach allows for more free speech while cutting down on the potential harm associated with virally abusive content.

The number of tweets containing hateful content on Twitter increased sharply in the week before Musk tweeted on November 23. According to the Center for Countering Digital Hate, the effects of hate speech, or views, were declining – an example from researchers hinting at the spread of such material, while Musk claims a reduction in visibility.

Researchers said tweets containing anti-Black words that week were triple the number seen in the month before Musk took office, while tweets containing gay slurs increased by 31%.

‘High risk, move fast’

Irwin, who joined the company in June and previously held security roles at other companies including Amazon.com and Google, shrugged off suggestions that Twitter does not have the resources or desire to protect the platform.

She said the layoffs did not significantly affect full-time employees or contractors working on what the company referred to as its “health” divisions, which include “critical areas” such as child safety and content moderation. Huh.

Two sources familiar with the cuts said more than 50% of the health engineering unit had been laid off. Irwin did not immediately respond to a request for comment on the claim, but previously denied that the health team was severely affected by the layoffs.

He added that the number of people working on child safety has not changed since the acquisition, and that the team’s product manager is still there. Irwin said Twitter back-filled some positions for people leaving the company, though he declined to provide specific figures for the extent of the turnover.

She said Musk was focusing on using automation more, arguing that the company had erred in the past on the side of using time- and labor-intensive human review of harmful material.

“He encouraged the team to take more risks, to go faster, to keep the stage safe,” she said.

On child safety, for example, Irwin said Twitter has shifted toward automatically taking down tweets reported by credible figures with a track record of automatically flagging harmful posts.

Carolina Cristofaletti, Threat Intelligence Researcher at TRM Labs, which specializes in child sexual abuse material, said she recently noticed on Twitter that some content is being taken down as fast as within 30 seconds of reporting it, or receiving a report. without confirmation of his decision.

In the interview on Thursday, Irwin said that Twitter has closed about 44,000 accounts involved in child safety breaches in cooperation with cybersecurity group Ghost Data.

Twitter is also restricting hashtags and search results often associated with abuse for the purpose of viewing “teenage” pornography. He said that previous concerns about the effect of such restrictions on the permitted uses of the terms were dispelled.

The use of “credible journalists” is “something we’ve discussed in the past on Twitter, but there was some hesitation and obviously some delay,” Irwin said.

“I think now we have the ability to really move forward with things like this,” she said.

Reporting by Katie Paul and Sheila Dang; Editing by Kenneth Lee and Anna Driver

Our Standards: Thomson Reuters Trust Doctrine.