Intermediary liability

safe harbor
Kyung Sin (“KS”) Park
Open Net/ Korea University
Director/Professor
kyungsinpark@korea.ac.kr
Civilizational significance of the Internet
• Giving powerless people the tools of mass communication
• What is mass communication? TV and newspaper (reaching multitude)
• TV/government, newspaper/corporations, also LIMITED SPACE  elite vs
people
• Formal democracy vs. substantive democracy
• Substantive democracy requires equality in communication
• Does internet give us that?
• 2012 Korean Constitutional Court on Internet real name law : “overcome
hierarchy offline in age, gender, social status”
• 2011 Korean Constitutional Court on election restriction: “online
communication requires AFFIRMATIVE conduct of receiver, so not easily
affected by financial dominance”
Intermediary Liability Safe Harbor
• ISPs/ Social media platforms/ Web hosts/ Search engines
• When should intermediaries be held liable for “aiding and abetting” online illegal
content?
• SAFE HARBOR: No liability as long as not aware of illegal contents, why?  if not,
GENERAL MONITORING or Prior Censorship  Internet becomes like TV and
newspaper subject to gate keeping  People lose the power of speaking to one
another without approval
• EU e-Commerce Directive Article 13-15
• US Digital Millenium Copyright Act section 512
• Japan Provider Liability Article 3 (1)-(2)
Notice and Takedown (Liability-Exempting): “shall NOT be liable if. . .”
not liable if you do X. Not doing X does not mean liability but just falls back on
ordinary torts  bright line rule of exemption for unknown contents
Three Types of intermediary liability
• Broad immunity: Communication Decency Act Section 230
• Intermediary Liability Safe Harbor
• Strict liability:
• Thailand’s Computer Crimes Act 2007 (CCA 2007) Article 14-15 criminal
sanctions imposed, inter alia, for allowing publication of information on
public computers in circumstances where the disseminated information is false
and is likely to cause damage to a third party or the country’s national security
 Chiranuch Premchaiporn
• CHINA – liability for failing to monitor, remove, even without notice
 Consideration should be given insulating intermediaries
from liability for content produced by others where
Intermediary liability should only be incurred if the intermediary has
specifically intervened in the content, which is published
liability safe online or fails to take down content following a court
order (contrary to the practice of notice and takedown).
harbor as
2011 Joint Declaration of UN, OAS, OSCE, and ACHPR on
international Freedom of Expression and the Internet, June 2011

standard [N]o one should be held liable for content on the internet of
which they are not the author. Indeed, no State should use or
force intermediaries to undertake censorship on its behalf.

Report of the UN Special Rapporteur on the promotion and

protection of the right to freedom of opinion and expression, 16 May
2011, A/HRC/17/27, para. 43.
International Soft Law
2016 Trump election  “fake news” debate
• German Network Enforcement Act (NetzDG) 2017
• Australia’s 2019 “abhorrent violent content” law
• France 2020 Avia law
Mandatory notice and takedown law
- Liability-imposing: “. . shall be liable if fails to take down within 7 days/24
hours/1 hr”
- Technically ok under SAFE HARBOR because applies only to NOTICED content
but. . .
Platform’s tendencies to err on the side of taking down vs. keeping it up
MANY lawful postings taken down
France Avia Law struck down– Constitutional
Conseil (June 2020)
• Part I: Terrorist content, child pornography AS notified by administrative authority – within 1 hr
• Part II: Other “manifestly illegal” content – within 24 hours of notice
• Unconstitutional b/c time too short, extrajudicial  criminal penalty not proportionate
• Probably considered impact on free speech – ”false positives”
• Conseil national du numerique (French Digital Commission), la Commission nationale
consultative des droits de l’homme (French Human Rights Commission) opposed.
• Notice that Part I is not even a liability law but direct administrative censorship.
• 2022 October Spanish Supreme Court on blocking order on womenonweb.org – administrative
censorship unconstitutional!
• 2009 June French Supreme Court on HADOPI law – administrative cut-off of internet access
unconstitutional
• 2014 Philippines Supreme Court – administrative censorship unconstitutional
We had it all along in Korea and Asia!
• Copyright Act – attempt to copy DMCA 512 but break into 2 sentences – “not
liable if take down” + “must take down if noticed”
• Network Act - ”must take down if noticed” (2007)
• -Missing ”liability-exempting” language and only “liability imposing”
• Problems of over-blocking – many legal contents taken down.
• Other Asian adaptations of mandatory notice and takedown:
• Malaysian Copyright Act (2012)
• Indonesian commerce platform circular (2016)
Vs. (liability-exempting)
• Japanese Provider Law (2001)
• Indian IT Act (2011)
However, Germany’s NetzDG (2017)
• Over 2 million registered users
• Take down illegal contents defined by Germany’s Criminal Code.
• flagged by individuals.
• “manifestly unlawful” : within 24 hours, all other “unlawful content”, within 7 days.
• Or face fines of up to 50 million euros
On face, no violation of international standard on safe harbor but exploits the
grey area by requiring “noticed” contents to be taken down if and when platform
operators are not aware of illegality.
“over-implementation” by providing an incentive to err on the side of caution rather than
free expression
• “privatizing” online censorship because of the scalability issue. No public
control but reliance on platforms’ decisions
 Spread of NetzDG in Broader Contents Scope
• 2019 Philippines Anti False Content Act – mentions NetzDG
• 2018 Malaysia Anti Fake News Act – mentions NetzDG
• 2018 Vietnam Cybersecurity law – “propaganda against Vietman”, etc., -
similar to NetzDG, e.g., 24 hours
• 2019 Singapore Protection from Online Falsehoods and Manipulation Act –
false statement of facts - mentions NetzDG
• 2021 Indonesia MR5 – “prohibited content”, e.g., 4 hours, 24 hours of
flagging by Ministry - similar to NetzDG
• 2022 Myanmar Cybersecurity Bill – must remove ”timely” all prohibited
content after flagging by the department including “complained of stmts
damaging another’s social standing and livelihood”
Moving forward
• Direct administrative censorship
• General monitoring obligation
• In the meantime. . .
• 2021 Australia Online Safety Act - eSafety Commissioner’s removal
notice to be acted upon
• 2021 Canada Online Harms bill – upload filtering, 24 hours
mandatory (private) notice and takedown.
• 2021 UK Online Safety bill – duty to detect and remove and Ofcom
able to penalize if not fulfilled
FOSTA (Fight Online Sex-Trafficking Act)
• CDA 230: no liability even for known contents.
• Exception from CDA 230: “knowingly assist[], support[], or facilitate[]” activity violating federal
sex trafficking law
• Difference b/w trafficking and voluntary prostitution
• the law likely placed sex workers at increased risk, as people lost access to websites used to
advertise and communicate about the sex trade and were forced to work on the street in
dangerous conditions.
• The law may even have hindered efforts to investigate and prosecute trafficking because as
websites such as Craiglist personal ads shut down, the trade were forced to find other venues
less transparent
• Also, a wide range of companies shut down or limited access to online services, including
services that had little or nothing to do with sex trafficking.
• Good Samaritan aspect of CDA 230: Moderation does not impose knowledge on you (also, no
liability for takedown – but already there in general law).
• “knowing” assistance = knowing illegality or knowing existence of contents?
Korea: Anti-”Nth Room” Law
• Nth Room: Forced sexual activities, not rape, using threats of disclosing nude
photos – broadcast over various chatrooms on Telegram for fee varying
depending on which “N”th room
• Law requires platforms to take “technical measures” to prevent circulation of
illegally filmed material
• Illegally filmed material – nonconsensual sexual video, sexual “deep fakes”,
”child pornography(CSAM)”,
• All videos uploaded in major apps/webs in Korea subject to filtering to
compare against administratively pre-curated database of ‘illegally filmed
material’
• General monitoring obligation? ”Function creep”?
Why change in 2016?
• Buzzfeed 1: : fake news gone viral (e.g., Pope Endorses Trump) beyond real news.
• Buzzfeed 2: 40% of Trump voters believing in Democrats’ child sex slave ring.” 36%
believing in Kenyan birth of Obama
 “Fake news is affecting world history!!!”
• “Fake News = NOT false news but news from FAKE SITES. Fake news is the offsprings of
digitalization.
• “’METANESS’ about fake news. The gravamen of fake news is not that the news is fake
but the source is fake (i.e. decorated like legit news site). The harm does not come from
the fact that people believe the story but they believe (incorrectly) that it was carried by
reputable media.
• “ once believed to have been picked up by reputable media, goes viral again not
because people believe the story but people believe the fact of coverage by reputable
media.  that alone does the magic e.g., casting a cloud of doubt on Hilary’s candidacy
My answer: Are fake news really a problem?
1. Do we know whether the stories were believed by people who
shared them on Facebook? Maybe, fake news were shared just for fun
not because the substance were believed. Look at Fake sites like
WorldPoliticus.com, ABCNews.com.co. NOT distinguishable from
supermarket tabloids (The National Enquirer, Star, The Globe, 
National Examiner) or “red tops” – (e.g. Alien Endorses Trump). Will
we regulate tabloids as well?
2. Harmful controversies believed by people (i.e. Obama’s Kenyan birth)
ARE NOT
FAKE NEWS shared through social media but false information shared
by POLITICIANS.