Beyond data protection
For the group assembled this week in Brussels for Computers, Privacy, and Data Protection, the General Data Protection Regulation that came into force in May 2018 represented the culmination of years of effort. The mood, however, is not so much self-congratulatory as "what's next?".
The first answer is a lot of complaints. An early panel featured a number of these. Max Schrems, never one to shirk, celebrated GDPR day in 2018 by joining with La Quadrature du Net to file two complaints against Google, WhatsApp, Instagram, and Facebook over "forced consent". Last week, he filed eight more complaints against Amazon, Apple, Spotify, Netflix, YouTube, SoundCloud, DAZN, and Flimmit regarding their implementation of subject access rights. A day or so later, the news broke: the French data protection regulator, CNIL, has fined Google €50 million (PDF) on the basis of their complaint - the biggest fine so far under the new regime that sets the limit at 4% of global turnover. Google is considering an appeal.
It's a start. We won't know for probably five years whether GDPR will have the intended effect of changing the balance of power between citizens and data-driven companies (even though one site is already happy to call it a failure already. Meanwhile, one interesting new development is Apple's crackdown on Facebook and then Google for abusing its enterprise app system to collect comprehensive data on end users. While Apple is certainly far less dependent on data collection than the rest of GAFA/FAANG, this action is a little like those types of malware that download anti-virus software to clean your system of the competition.
The second - more typical of a conference - is to stop and think: what doesn't GDPR cover? The answers are coming fast: AI, automated decision-making, household or personal use of data, and (oh, lord) blockchain. And, a questioner asked late on Wednesday, "Is data protection privacy, data, or fairness?"
Several of these areas are interlinked: automated decision-making is currently what we mean when we say "AI", and while we talk a lot about the historical bias stored in data and the discrimination that algorithms derive from training data and bake into their results. Discussions of this problem, Angsar Koene tend to portray accuracy and fairness as a tradeoff, with accuracy presented as a scientifically neutral reality and fairness as a fuzzy human wish. Instead, he argued, accuracy depends on values we choose to judge it by. Why shouldn't fairness just be one of those values?
A bigger limitation - which we've written about here since 2015 - is that privacy law tends to focus on the individual. Seda Gürses noted that focusing on the algorithm - how to improve it and reduce its bias - similarly ignores the wider context and network externalities. Optimize the Waze algorithm so each driver can reach their destination in record time, and the small communities whose roads were not built for speedy cut-throughs bear the costs of extra traffic, noise, and pollution they generate. Next-generation privacy will have to reflect that wider context; as Dennis Hirsch put it, social protection rather than individual control. As Schrems' and others' complaints show, individual control is rarely ours on today's web in any case.
Privacy is not the only regulation that suffers from that problem. At Tuesday's pre-conference Privacy Camp, several speakers deplored the present climate in which platforms' success in removing hate speech, terrorist content, and unauthorized copyright material is measured solely in numbers: how many pieces, how fast. Such a regime does not foster thoughtful consideration, nuance, respect for human rights, or the creation of a robust system of redress for the wrongly accused. "We must move away from the idea that illegal content can be perfectly suppressed and that companies are not trying hard enough if they aren't doing it," Mozilla Internet policy manager Owen Bennett said, going on to advocate for a wider harm reduction approach.
The good news, in a way, is that privacy law has fellow warriors: competition, liability, and consumer protection law. The first two of those, said Mireille Hildebrandt need to be rethought, in part because some problems will leave us no choice. She cited, for example, the energy market: as we are forced to move to renewables both supply and demand will fluctuate enormously. "Without predictive technology I don't see how we can solve it." Continuously predicting the energy use of each household will, she wrote in a paper in 2013 (PDF), pose new threats to privacy, data protection non-discrimination, and due process.
One of the more interesting new (to me, at least) players on this scene is Algorithm Watch, which has just released a report on algorithmic decision-making in the EU that recommends looking at other laws that are relevant to specific types off decisions, such as applying equal pay legislation to the gig economy. Data protection law doesn't have to do it all.
Some problems may not be amenable to law at all. Paul Nemitzposed this question: given that machine learning training data is always historical, and that therefore the machines are always perforce backward-looking, how do we as humans retain the drive to improve if we leave all our decisions to machines? No data protection law in the world can solve that.
Illustrations: The CPDP 2019 welcome sign in Brussels.
Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. Stories about the border wars between cyberspace and real life are posted occasionally during the week at the net.wars Pinboard - or follow on Twitter.