Twitter said more than one-half of tweets it took action on for abuse during the first half of 2019 were surfaced using its proprietary artificial intelligence technology, lessening its reliance on users reporting those offenses, adding that just 20% of those tweets were caught by its systems during the same period last year.
There was also a 105% leap in accounts locked or suspended for violating the Twitter rules compared with the first half of 2018, which Twitter attributed to a combination of its focus on proactively surfacing potentially violative content for human review and its inclusion of impersonation data.
The social network released its 15th biannual Transparency Report this week, providing transparency on global trends in enforcement on the social network.
Twitter provided some color on specific policy areas.
The social network saw a 48% year-over-year increase in accounts reported for potential violations of its private information policies, suspending 119% more accounts for these offenses than it did during the second half of 2018. Twitter wrote, “This increase may be attributed to the launch of improvements to our reporting flow that make it easier to report private information, as well as changes to our internal enforcement processes that permit bystanders to report potential private information violations for review.”
There was a 37% increase in accounts reported for potential violations of Twitter’s policies on sensitive media, with 41% more accounts subject to actions.
Accounts reported for potential violations of the social network’s hateful conduct policies were up 48%, with actions taken on 133% more accounts than during the second half of 2018.
Action was taken on 68% more accounts for abuse than in the last reporting period, with a 22% spike in accounts reported for potential violations of those policies.
During the first half of 2019, enforcement actions were taken against 124,339 accounts for violating Twitter’s impersonation policy.
Twitter also provided an update on its anti-spam efforts, saying that during the first half of the year, anti-spam challenges—where users are asked to provide phone numbers or email addresses, or to fill in ReCAPTCHA codes, in order to verify that the account is being controlled by a human—were down nearly 50%.
The social network said 115,861 accounts were suspended for violations related to the promotion of terrorism during the first half of 2019, down 30% from the second half of 2018, with 87% of those accounts proactively flagged using internal, proprietary removal tools.
And 244,188 accounts were suspended for violations related to child sexual exploitation, with 91% of those surfaced by a combination of technology solutions, including PhotoDNA and internal, proprietary tools.
Twitter said it saw a 101% increase in Digital Millennium Copyright Act takedown notices from the prior reporting period, but many were incomplete or not actionable, with the social network citing high volumes of those from Turkey, Japan and Brazil.
Trademark notices were up 39% in the first half of 2019 versus the second half of 2018.
The U.S. continued to account for the largest portion of legal requests for account information, at 29% during this reporting period.
Twitter said people affected by three National Security Letters were notified during the first half of 2019 after gag orders were lifted, adding that nondisclosure orders for 17 total NSLs have been lifted to date.
The social network received roughly two-thirds more legal requests to remove content in the first half of the year compared with the second half of last year, with 80% of those requests coming from Japan, Russia and Turkey.
Content was withheld in a country 2,457 times during the period, at either the account or tweet level.