Human rights in the age of

Direct Threats: Internet companies that host content use AI to flag posts that violate their terms of service. 
Governments exerting formal and informal pressure on companies to address the problem of alleged 
terrorist content, hate speech, and so-called “fake news,” but without clear standards or definitions, has 
led to increased use of automated systems.
77
A law recently passed in Germany requires social media sites 
to remove a wide range of content within 24 hours after it has been flagged (or up to seven days in cases 
that are less clear-cut).
78
Because AI is imperfect and companies are pressured to take down questionable 
content so quickly, much of the content is removed in error.
79
YouTube removed more than 100,000 videos 
documenting atrocities in Syria after they were flagged, . These videos often serve as the only evidence of 
horrific crimes and human rights violations, and YouTube’s policy carves out exceptions for violent content 
when it is of important educational or documentary value. 
80
Yet they were still taken down.
Authoritarian governments can use similar technology to increase censorship. The Chinese government is 
already replacing some of its human censors with AI. Popular Chinese video platform iQiyi uses ML to identify 
pornographic and violent content, as well as content deemed “politically sensitive.” Because ML cannot deal with 
nuance, flagged content is currently reviewed by humans, though this may change as the technology becomes 
more sophisticated and industry sees the human resources required for review as an unnecessary expense.
81
76 Article 19 of the UDHR and Article 19 of the ICCPR; Article 18 of the ICCPR and UDHR, Articles 21 and 22 of the ICCPR, Article 20 of the UDHR
77 A Freedom House survey found 30 of 65 of governments attempted to control online discussions. https://freedomhouse.org/article/new-report-
freedom-net-2017-manipulating-social-media-undermine-democracy.
78 “Germany starts enforcing hate speech law,” The BBC, January 1, 2018, https://www.bbc.com/news/technology-42510868.
79 Denis Nolasco and Peter Micek, “Access Now responds to Special Rappoteur Kaye on ‘Content Regulation in the Digital Age, Access Now, January 11, 2018 
80 Kate O’Flaherty, “YouTube keeps deleting evidence of Syrian chemical weapon attacks,” Wired, June 26, 2018, http://www.wired.co.uk/article/
chemical-weapons-in-syria-youtube-algorithm-delete-video.
81 Yuan Tang, “Artificial intellifence takes jobs from Chinese censors,” Financial Times, May 21, 2018, https://www.ft.com/content/9728b178-59b4-11e8-
bdb7-f6677d2e1ce8.

---

accessnow.org
23