How a small French privacy ruling could remake adtech for good
A ruling in late October against a little-known French adtech firm that popped up on the national data watchdog’s website earlier this month is causing ripples of excitement to run through privacy watchers in Europe who believe it signals the beginning of the end for creepy online ads.
The excitement is palpable.
Impressively so, given the dry CNIL decision against mobile “demand side platform” Vectaury was only published in the regulator’s native dense French legalese.
The CMP fails exactly as you would expect:— Robin Berjon (@robinberjon) November 16, 2018
1) The consent is not informed;
2) The consent is not specific;
3) The consent is not affirmative.
Given the high-risk nature of the processing and its opacity, this isn't even a very strict interpretation.
Digital advertising trade press AdExchanger picked up on the decision yesterday.
Here’s the killer paragraph from CNIL’s ruling — translated into “rough English” by my TC colleague Romain Dillet:
The requirement based on the article 7 above-mentioned isn’t fulfilled with a contractual clause that guarantees validly collected initial consent. The company VECTAURY should be able to show, for all data that it is processing, the validity of the expressed consent.
In plainer English, this is being interpreted by data experts as the regulator stating that consent to processing personal data cannot be gained through a framework arrangement which bundles a number of uses behind a single “I agree” button that, when clicked, passes consent to partners via a contractual relationship.
CNIL’s decision suggests that bundling consent to partner processing in a contract is not, in and of itself, valid consent under the European Union’s General Data Protection Regulation (GDPR) framework.
Consent under this regime must be specific, informed and freely given. It says as much in the text of GDPR.
But now, on top of that, the CNIL’s ruling suggests a data controller has to be able to demonstrate the validity of the consent — so cannot simply tuck consent inside a contractual “carpet-bag” that gets passed around to everyone else in their chain as soon as the user clicks “I agree.”
This is important, because many widely used digital advertising consent frameworks rolled out to websites in Europe this year — in claimed compliance with GDPR — are using a contractual route to obtain consent, and bundling partner processing behind often hideously labyrinthine consent flows.
The experience for web users in the EU right now is not great. But it could be leading to a much better internet down the road.
Where’s the consent for partner processing?
Even on a surface level the current crop of confusing consent mazes look problematic.
But the CNIL ruling suggests there are deeper and more structural problems lurking and embedded within. And as regulators dig in and start to unpick adtech contradictions it could force a change of mindset across the entire ecosystem.
As ever, when talking about consent and online ads the overarching point to remember is that no consumer given a genuine full disclosure about what’s being done with their personal data in the name of behavioral advertising would freely consent to personal details being hawked and traded across the web just so a bunch of third parties can bag a profit share.
This is why, despite GDPR being in force (since May 25), there are still so many tortuously confusing “consent flows” in play.
The longstanding online T&Cs trick of obfuscating and socially engineering consent remains an unfortunately standard playbook. But, less than six months into GDPR we’re still very much in a “phoney war” phase. More regulatory rulings are needed to lay down the rules by actually enforcing the law.
And CNIL’s recent activity suggests more to come.
In the Vectaury case, the mobile ad firm used a template framework for its consent flow that had been created by industry trade association and standards body, IAB Europe.
It did make some of its own choices, using its own wording on an initial consent screen and pre-ticking the purposes (another big GDPR no-no). But the bundling of data purposes behind a single opt in/out button is the core IAB Europe design. So CNIL’s ruling suggests there could be trouble ahead for other users of the template.
IAB Europe’s CEO, Townsend Feehan, told us it’s working on a statement reaction to the CNIL decision, but suggested Vectaury fell foul of the regulator because it may not have implemented the “Transparency & Consent Framework-compliant” consent management platform (CMP) framework — as it’s tortuously known — correctly.
So either “the ‘CMP’ that they implemented did not align to our Policies, or choices they could have made in the implementation of their CMP that would have facilitated compliance with the GDPR were not made,” she suggested to us via email.
Though that sidesteps the contractual crux point that’s really exciting privacy advocates — and making them point to the CNIL as having slammed the first of many unbolted doors.
The French watchdog has made a handful of other decisions in recent months, also involving geolocation-harvesting adtech firms, and also for processing data without consent.
So regulatory activity on the GDPR+adtech front has been ticking up.
Its decision to publish these rulings suggests it has wider concerns about the scale and privacy risks of current programmatic ad practices in the mobile space than can be attached to any single player.
So the suggestion is that just publishing the rulings looks intended to put the industry on notice…
Other factor: the technicity and opacity of ad bidding systems is used to justify greater transparency requirements. People cannot expect to consent to what they don't understand (or even know exist), therefore the adtech ecosystem is de facto under *stronger* requirements.— Robin Berjon (@robinberjon) November 16, 2018
Meanwhile, adtech giant Google has also made itself unpopular with publisher “partners” over its approach to GDPR by forcing them to collect consent on its behalf. And in May a group of European and international publishers complained that Google was imposing unfair terms on them.
The CNIL decision could sharpen that complaint too — raising questions over whether audits of publishers that Google said it would carry out will be enough for the arrangement to pass regulatory muster.
This means that if someone gains consent for you, and you have a contract saying it's their responsibility to do so, you *still* have the obligation to verify that the consent is valid.— Robin Berjon (@robinberjon) November 16, 2018
For a demand-side platform like Vectaury, which was acting on behalf of more than 32,000 partner mobile apps with user eyeballs to trade for ad cash, achieving GDPR compliance would mean either asking users for genuine consent and/or having a very large number of contracts on which it’s doing actual due diligence.
Yet Google is orders of magnitude more massive, of course.
The Vectaury file gives us a fascinating little glimpse into adtech “business as usual.” A business which also wasn’t, in the regulator’s view, legal.
The firm was harvesting a bunch of personal data (including people’s location and device IDs) on its partners’ mobile users via an SDK embedded in their apps, and receiving bids for these users’ eyeballs via another standard piece of the programmatic advertising pipe — ad exchanges and supply side platforms — which also get passed personal data so they can broadcast it widely via the online ad world’s real-time bidding (RTB) system. That’s to solicit potential advertisers’ bids for the attention of the individual app user… The wider the personal data gets spread, the more potential ad bids.
That scale is how programmatic works. It also looks horrible from a GDPR “privacy by design and default” standpoint.
The sprawling process of programmatic explains the very long list of “partners” nested non-transparently behind the average publisher’s online consent flow. The industry, as it is shaped now, literally trades on personal data.
So if the consent rug it’s been squatting on for years suddenly gets ripped out from underneath it, there would need to be a radical reshaping of ad-targeting practices to avoid trampling on EU citizens’ fundamental right.
GDPR’s really big change was supersized fines. So ignoring the law would get very expensive.
Oh hai real-time bidding!
In Vectaury’s case, CNIL discovered the company was holding the personal data of a staggering 67.6 million people when it conducted an on-site inspection of the company in April 2018.
That already sounds like A LOT of data for a small mobile adtech player. Yet it might actually have been a tiny fraction of the personal data the company was routinely handling — given that Vectaury’s own website claims 70 percent of collected data is not stored.
In the decision, there was no fine, but CNIL ordered the firm to delete all data it had not already deleted (having judged collection illegal given consent was not valid), and to stop processing data without consent.
But given the personal-data-based hinge of current-gen programmatic adtech, that essentially looks like an order to go out of business. (Or at least out of that business.)
And now we come to another interesting GDPR adtech complaint that’s not yet been ruled on by the two DPAs in question (Ireland and the U.K.) — but which looks even more compelling in light of the CNIL Vectaury decision because it picks at the adtech scab even more daringly.
Filed last month with the Irish Data Protection Commission and the U.K.’s ICO, this adtech complaint — the work of three individuals, Johnny Ryan of private web browser Brave; Jim Killock, exec director of digital and civil rights group, the Open Rights Group; and University College London data protection researcher, Michael Veale — targets the RTB system itself.
Here’s how Ryan, Killock and Veale summarized the complaint when they announced it last month:
Every time a person visits a website and is shown a “behavioural” ad on a website, intimate personal data that describes each visitor, and what they are watching online, is broadcast to tens or hundreds of companies. Advertising technology companies broadcast these data widely in order to solicit potential advertisers’ bids for the attention of the specific individual visiting the website.
A data breach occurs because this broadcast, known as an “bid request” in the online industry, fails to protect these intimate data against unauthorized access. Under the GDPR this is unlawful.
The GDPR, Article 5, paragraph 1, point f, requires that personal data be “processed in a manner that ensures appropriate security of the personal data, including protection against unauthorised or unlawful processing and against accidental loss.” If you can not protect data in this way, then the GDPR says you can not process the data.
Ryan tells TechCrunch that the crux of the complaint is not related to the legal basis of the data sharing but rather focuses on the processing itself — arguing “that it itself is not adequately secure… that there aren’t adequate controls.”
Though he says there’s a consent element too, and so sees the CNIL ruling bolstering the RTB complaint. (On that keep in mind that CNIL judged Vectaury should not have been holding the RTB data of 67.6M people because it did not have valid consent.)
“We do pick up on the issue of consent in the complaint. And this particular CNIL decision has a bearing on both of those issues,” he argues. “It demonstrates in a concrete example that involved investigators going into physical premises and checking the machines — it demonstrates that even one small company was receiving tens of millions of people’s personal data in this illegal way.
“So the breach is very real. And it demonstrates that it’s not unreasonable to suggest that the consent is meaningless in any case.”
Reaching for a handy visual explainer, he continues: “If I leave a briefcase full of personal data in the middle of Charing Cross station at 11am and it’s really busy, that’s a breach. That would have been a breach back in the 1970s. If my business model is to drive up to Charing Cross station with a dump-truck and dump briefcases onto the street at 11am in the full knowledge that my business partners will all scramble around and try and grab them — and then to turn up at 11.01am and do the same thing. And then 11.02am. And every microsecond in between. That’s still a fucking data breach!
“It doesn’t matter if you think you’ve consent or anything else. You have to [comply with GDPR Article 5, paragraph 1, point f] in order to even be able to ask for a legal basis. There are plenty of other problems but that’s the biggest one that we highlighted. That’s our reason for saying this is a breach.”
“Now what CNIL has said is this company, Vectaury, was processing personal data that it did not lawfully have — and it got them through RTB,” he adds, spelling the point out. “So back to the GDPR — GDPR is saying you can’t process data in a way that doesn’t ensure protection against unauthorized or unlawful processing.”
In other words, RTB as a funnel for processing personal data looks to be on inherently shaky ground because it’s inherently putting all this personal data out there and at risk…
What’s bad for data brokers…
In another loopback, Ryan says the regulators have been in touch since their RTB complaint was filed to invite them to submit more information.
He says the CNIL Vectaury decision will be incorporated into further submissions, predicting: “This is going to be bounced around multiple regulators.”
The trio is keen to generate extra bounce by working with NGOs to enlist other individuals to file similar complaints in the other EU Member States — to make the action a pan-European push, just like programmatic advertising itself.
“We now have the opportunity to connect our complaint with the excellent work that Privacy International has done, showing where these data end up, and with the excellent work that CNIL has done showing exactly how this actually applies. And this decision from CNIL takes, essentially my report that went with our complaint and shows exactly how that applies in the real world,” he continues.
“I was writing in the abstract — CNIL has now made a decision that is very much not in the abstract, it’s in the real world affecting millions of people… This will be a European-wide complaint.”
But what does programmatic advertising that doesn’t entail trading on people’s grubbily obtained personal data actually look like? If there were no personal data in bid requests Ryan believes quite a few things would happen. Such as, for e.g. the demise of clickbait.
“There would be no way to take your TechCrunch audience and buy it cheaper on some shitty website. There would be no more of that arbitrage stuff. Clickbait would die! All that nasty stuff would go away,” he suggests.
(And, well, full disclosure: We are TechCrunch — so we can confirm that does sound really great to us!)
He also reckons ad values would go up. Which would also be good news for publishers. (“Because the only place you could buy the TechCrunch audience would be on TechCrunch — that’s a really big deal!”)
He even suggests ad fraud might shrink because the incentives would shift. Or at least they could so long as the “worthy” publishers that are able to survive in the new ad world order don’t end up being complicit with bot fraud anyway.
As it stands, publishers are being screwed between the twin plates of the dominant adtech platforms (Google and Facebook), where they are having to give up a majority of their ad revenue — leaving the media industry with a shrinking slice of ad revenues (that can be as lean as ~30 percent).
That then has a knock on impact on funding newsrooms and quality journalism. And, well, on the wider web too — given all the weird incentives that operate in today’s big tech social media platform-dominated internet.
While a privacy-sucking programmatic monster is something only shadowy background data brokers that lack any meaningful relationships with the people whose data they’re feeding the beast could truly love.
And, well, Google and Facebook.
Ryan’s view is that the reason an adtech duopoly exists boils down to the “audience leakage” being enabled by RTB. Leakage which, in his view, also isn’t compliant with EU privacy laws.
He reckons the fix for this problem is equally simple: Keep doing RTB but without any personal data.
A real-time ad bidding system that’s been stripped of personal data does not means no targeted ads. It could still support ad targeting based on real-time factors such as an approximate location (say to a city region) and/or generic and aggregated data.
Crucially it would not use unique identifiers that enable linking ad bids to an individual’s entire digital footprint and bid request history — as is the case now. Which essentially translates into RIP privacy rights.
Ryan argues that RTB without personal data would still offer plenty of “value” to advertisers — who could still reach people based on general locations and via real-time interests. (It’s a model that sounds much like what privacy search engine DuckDuckGo is doing, and also been growing.)
The really big problem, though, is turning the behavioural ad tanker around. Given that the ecosystem is embedded, even as the duopoly milks it.
That’s also why Ryan is so hopeful now, though, having parsed the CNIL decision.
His reading is regulators will play a decisive role in pushing the ad industry’s trigger — and force through a much-needed change in their targeting behaviour.
“Unless the entire industry moves together, no one can be the first to remove personal data from bid requests but if the regulators step in in a big way… and say you’re all going to go out of business if you keep putting personal data into bid requests then everyone will come together — like the music industry was forced to eventually, under Steve Jobs,” he argues. “Everyone can together decide on a new short-term disadvantageous but long term highly advantageous change.”
Of course, such a radical reshaping is not going to happen overnight. Regulatory triggers tend to be slow motion unfoldings at the best of times. You also have to factor in the inexorable legal challenges.
But look closely and you’ll see both momentum massing behind privacy — and regulatory writing on the wall.
“Are we going to see programmatic forced to be non-personal and therefore better for every single citizen of the world (except, say, if they work for a data broker),” adds Ryan, posing his own concluding question. “Will that massive change, which will help society and the web… will that change happen before Christmas? No. But it’s worth working on. And it’s going to take some time.
“It could be two years from now that we have the finality. But a finality there will be. Detroit was only able to fight against regulation for so long. It does come.”
Who’d have thought “taking back control” could ever sound so good?