Media coverage
Learning from Europe but looking beyond for privacy law By Jules Polonetsky, Opinion Contributor — 10/16/18 06:35 PM EDT
The views expressed by contributors are their own and not the view of The Hill
© Getty Images
Not a day seems to pass without reports of a new data breach or privacy miss-step by a
company. It’s no surprise that the White House, Congress, industry and civil society are
increasingly in agreement about the need for comprehensive federal privacy legislation.
What should such a privacy law look like? This week the Senate Commerce Committee held
a hearing seeking lessons from the recently adopted European General Data Protection Law,
a detailed framework governing every aspect of collection and use of personal data. There is
value in looking to the basic principles of the GDPR, but there are also areas where the
GDPR principles can be refined to most effectively support the rights and freedoms U.S.
legislators want to protect.
In a world of international data flows, national and global interoperability is critical. The
GDPR seeks to have one set of privacy rules that are interpreted in a uniform way throughout
the continent, as European Data Protection Board Chair Andrea Jelinek explained in her
Senate testimony. Federal legislation that took some important lessons from GDPR would be
a step towards trans-Atlantic interoperability.
The GDPR is technology neutral and covers every type of company and business model. At a
time when data is collected across platforms—on the web, on mobile, with wearables, smart
home devices, and phone and facial tracking — emulating the GDPR’s comprehensive model
with a law that sets one clear set of rules will be easier to enforce and easiest for consumers
to understand.
Another central pillar of the GDPR is fairness, which means using data only as people would
reasonably expect and not using it in ways that have unjustified adverse effects, a concept
very similar to the current Federal Trade Commission statute authorizing legal action against
deceptive or unfair practices.
But there are many aspects of the GDPR that are less suitable for U.S. legislation. The GDPR
is comprehensive, written to regulate not only business, but also governments, political
campaigns and not for profits. The laws regulating government collection and use of data in
the United States do indeed need to be updated, but it seems unlikely that policymakers are
ready to address those issues in this bill. And given the concerns about the political collection
and use of data by Cambridge Analytica, a hard look at political campaign data uses is
warranted, but may face First Amendment protections in the U.S.
When it comes to children’s privacy, the U.S. children’s privacy law, COPPA, requires
parental consent for all collection of children’s personal data, with very limited exceptions.
The GDPR sets an age of 16, higher than 13 under COPPA, but only for the cases where
consent is required and only for providing “information society services.” In many cases, the
legitimate interest and opt-out model is permitted for collecting kids data in Europe. COPPA
almost always requires parental consent and covers all online services, often with strict
verification requirements.
The GDPR also poses some challenges for AI and machine learning, since it specifies that
personal data must be collected only for a specified purpose, must be deleted or minimized
when not needed for that purpose. While there are promising efforts to minimize data
collection for machine learning by conducting processes locally on user devices instead of
sending data back to a company, for many such uses of data today, large and representative
data sets are required to power new models, to ensure accuracy and to avoid bias. A U.S.
framework would be wise to ensure that uses of data for machine learning are supported,
when conducted responsibly.
U.S. law should also provide more flexibility for research than the GDPR, which mandates
that researchers provide more specificity about their plans and in many cases requires
continual permissions from individuals for future uses. Researchers often do not know what
type of insights a study will reveal, and rely on data sets that have been collected by third
parties from participants they cannot contact.
Any legislation should also consider the increasingly sophisticated privacy tools that are
emerging, including differential privacy to measure privacy risk, homomorphic encryption
that can enable privacy safe data analysis, and many new privacy compliance tools that are
helping companies better manage data. A law that will stand the test of time and successfully
protect privacy rights while enabling valuable uses of data should include mechanisms to
incentivize such technology measures.
So let’s look to lessons from GDPR for a federal bill, but also to the best ideas that privacy
experts in civil society and in industry can offer to develop a framework that prevents harms
while supporting responsible uses of data.
Jules Polonetsky