Privacy

Twitter faces privacy scrutiny from EU watchdogs after Mudge report

Comment

illustration of twitter logo, padlock pattern and shields
Image Credits: Bryce Durbin / TechCrunch

The explosive Twitter whistleblower complaint that was made public yesterday — detailing a raft of damning allegations across security, privacy and data protection issues (among others) by Twitter’s former former head of security, Peiter “Mudge” Zatko — contained references to European regulators along with claims that the social media firm had misled or intended to mislead regional oversight bodies over its compliance with local laws.

Two national data protection authorities in the EU, in Ireland and France, have confirmed to TechCrunch that they are following up on the whistleblower complaint.

Ireland, which is Twitter’s lead supervisor for the bloc’s General Data Protection Regulation (GDPR) — and previously led a GDPR investigation of a separate security incident that resulted in a $550,000 fine for Twitter — said it is “engaging” with the company in the wake of the publicity around the complaint.

“We became aware of the issues when we read the media stories [yesterday] and have engaged with Twitter on the matter,” the regulator’s deputy commissioner, Graham Doyle, told us.

While France’s DPA said it is investigating allegations made in the complaint.

“The CNIL is currently investigating the complaint filed in the U.S. For the moment we are not in a position to confirm or deny the accuracy of the alleged breaches,” a spokesperson for the French watchdog told us. “If the accusations are true, the CNIL could carry out checks that could lead to an order to comply or a sanction if breaches are found. In the absence of a breach, the procedure would be terminated.

Machine learning concerns

Ireland’s Data Protection Commission (DPC) and France’s national equivalent, the CNIL, were both cited in the ‘Mudge report’ — in one instance in relation to Zatko’s suspicion that Twitter intended to mislead them in relation to enquiries about data-sets used to train its machine learning algorithms in a similar way to how the complaint alleges Twitter misled the FTC years earlier over the issue.

In a section of the complaint given the title “misleading regulators in multiple countries”, Zatko asserts that the FTC had asked Twitter questions about the training material used to build its machine learning models.

“Twitter realized that truthful answers would implicate the company in extensive copyright / intellectual property violations,” runs the complaint, before asserting that Twitter’s strategy (which he says executives “explicitly acknowledged was deceptive”) was to decline to provide the FTC with the requested training material and instead point it to “particular models that would not expose Twitter’s failure to acquire appropriate IP rights”.

The two European regulators come into the picture because Zatko suggests they were poised to make similar enquiries this year — and he says he was told by a Twitter staffer that the company intended to try to use the same tactic it had deployed in response to earlier FTC enquiries on the issue, to derail regulatory scrutiny.

“In early 2022, the Irish-DPC and French-CNIL were expected to ask similar questions, and a senior privacy employee told Mudge that Twitter was going to attempt the same deception,” the complaint states. “Unless circumstances have changed since Mudge was fired in January, then Twitter’s continued operation of many of its basic products is most likely unlawful and could be subject to an injunction, which could take down most or all of the Twitter platform.”

Neither the Irish nor French watchdog responded to questions about the specific claims being made. So it’s not clear what enquiries the EU data protection agencies may have made — or be planning to make — of Twitter in relation to its machine learning training data-sets.

One possibility — and perhaps the most likely one, given EU data protection law — could be they have concerns or suspicions that Twitter processed personal data to build its AI models without having a proper legal basis for the processing.

In a separate example, the controversial facial recognition firm, Clearview AI, has in recent months faced a raft of regional enforcements from DPAs linked to its use of personal data for training its facial recognition models. Although the personal data in that case — selfies/facial biometrics — is among the most protected ‘sensitive’ class of data under EU law, meaning it carries the strictest requirements for legal processing (and it’s not clear whether Twitter might have been using similarly sensitive data-sets for training its AI models).

Selfie scraping Clearview AI hit with another €20M ban order in Europe

Cookies out of control?

The Mudge complaint also makes a direct claim that Twitter misled the CNIL over a separate issue — related to improper separation of cookie functions — after the French watchdog ordered it to amend its processes to come into compliance with relevant laws in December 2021.

Zatko alleges that up until Q2/Q3 of 2021 Twitter lacked sufficient understanding of how it was deploying cookies and what they were used for — and also that Twitter cookies were being used for multiple functions, such as ad tracking and security sessions.

“It was apparent Twitter was in violation of international data requirements across many regions of the world,” the complaint asserts.

A key tenet of European Union data protection law that applies here is ‘purpose limitation’ — i.e. the principle that personal data must be used for the stated (legitimate) purpose it was collected for; and that uses for data should not be bundled. So if Twitter was mingling cookie function for distinctly different purposes, such as marketing and security — as the complaint claims — that would create clear legal problems for it in the EU.

According to the complaint, the CNIL got wind of a cookie function problem at Twitter and ordered the company to fix at the end of last year, presumably relying on its competence under the EU’s ePrivacy Direction (which regulates use of tracking technologies like cookies).

Zatko writes that a new privacy engineering team at Twitter had worked “tirelessly” to disentangle cookie function in order to permit “some form of user choice and control” — to, for example, deny tracking cookies but accept security-related cookies — as would be required under EU law. And he says this fix was rolled out, exclusively in France, on December 31, 2021, but was immediately rolled back and disabled after Twitter encountered a problem — an ops SNAFU he seizes on to heap more blame on Twitter for failing to have a separate testing environment.

But while he writes that the bug was fixed “in a matter of hours”, he claims Twitter product and legal decision-makers blocked rolling it out for another month — until January 31, 2021 — “in order to extract maximum profit from French users before rolling out the fix”.

“Mudge challenged executives to claim this was anything other than an effort to prioritize incremental profits over user privacy and legal data privacy requirements,” the complaint also asserts, adding: “The senior leaders in that meeting confessed that Mudge was correct.”

Zatko makes a further claim that Twitter launched “proactive” legal action — in which he says they were “attempting to claim that all cookies were by definition critical and required, because the platform is powered by advertisements” — before going on to allege that during internal conversations he heard product staff stating the argument was “false and made in bad faith”.

Twitter was contacted for a response to the specific claims referenced in cited portions of the whistleblower’s report but at the time of writing it had not responded. But the company put out a general response to the Mudge report yesterday — dismissing the complaint as a “false narrative” by a disgruntled former employee, which it also claimed was “riddled with inconsistencies and inaccuracies”.

Regardless, the whistleblower complaint is already sparking fresh regulatory scrutiny of Twitter’s claims.

It’s not clear what penalties the company could face in the EU if regulators decide — on closer inspection — that it has breached regional requirements after following up on Mudge’s complaint.

The GDPR allows for penalties that scale up to 4% of annual global turnover — although Twitter’s prior GDPR penalty, for a separate security-related breach, fell far short of that. However enforcements are supposed to factor in the scale and extent (and indeed intent) of any violations — and the extensive failings being alleged by Mudge, could — if stood up by formal regulatory investigation — lead, eventually, to a far more substantial penalty.

The ePrivacy Directive, which gives CNIL competency to regulate Twitter’s cookies, empowers DPAs to issue “effective, proportionate and dissuasive” sanctions — so it’s hard to predict what that might mean in hard financial terms if it deems a fine is justified. But in recent years the French watchdog has issues a series of multi-million dollar fines to tech giants for cookie-related failures.

This includes two beefy penalties for Google — a $170 million fine in January over deceptive cookie consent banners; and a separate $120M fine in December 2020 for dropping tracking cookies without consent — as well as a $68 million fine for Facebook back in January (also for deceptive cookies), and a $42 million fine for Amazon at the end of 2020, also for dropping tracking cookies without consent.

Update: Twitter declined to provide public comment.

France spanks Google $170M, Facebook $68M over cookie consent dark patterns

Twitter fined ~$550K over a data breach in Ireland’s first major GDPR decision

More TechCrunch

Replacing Sutskever is Jakub Pachocki, OpenAI’s director of research.

Ilya Sutskever, OpenAI co-founder and longtime chief scientist, departs

Intuitive Machines made history when it became the first private company to land a spacecraft on the moon, so it makes sense to adapt that tech for Mars.

Intuitive Machines wants to help NASA return samples from Mars

As Google revamps itself for the AI era, offering AI overviews within its search results, the company is introducing a new way to filter for just text-based links. With the…

Google adds ‘Web’ search filter for showing old-school text links as AI rolls out

Blue Origin’s New Shepard rocket will take a crew to suborbital space for the first time in nearly two years later this month, the company announced on Tuesday.  The NS-25…

Blue Origin to resume crewed New Shepard launches on May 19

This will enable developers to use the on-device model to power their own AI features.

Google is building its Gemini Nano AI model into Chrome on the desktop

It ran 110 minutes, but Google managed to reference AI a whopping 121 times during Google I/O 2024 (by its own count). CEO Sundar Pichai referenced the figure to wrap…

Google mentioned ‘AI’ 120+ times during its I/O keynote

Firebase Genkit is an open source framework that enables developers to quickly build AI into new and existing applications.

Google launches Firebase Genkit, a new open source framework for building AI-powered apps

In the coming months, Google says it will open up the Gemini Nano model to more developers.

Patreon and Grammarly are already experimenting with Gemini Nano, says Google

As part of the update, Reddit also launched a dedicated AMA tab within the web post composer.

Reddit introduces new tools for ‘Ask Me Anything,’ its Q&A feature

Here are quick hits of the biggest news from the keynote as they are announced.

Google I/O 2024: Here’s everything Google just announced

LearnLM is already powering features across Google products, including in YouTube, Google’s Gemini apps, Google Search and Google Classroom.

LearnLM is Google’s new family of AI models for education

The official launch comes almost a year after YouTube began experimenting with AI-generated quizzes on its mobile app. 

Google is bringing AI-generated quizzes to academic videos on YouTube

Around 550 employees across autonomous vehicle company Motional have been laid off, according to information taken from WARN notice filings and sources at the company.  Earlier this week, TechCrunch reported…

Motional cut about 550 employees, around 40%, in recent restructuring, sources say

The keynote kicks off at 10 a.m. PT on Tuesday and will offer glimpses into the latest versions of Android, Wear OS and Android TV.

Google I/O 2024: Watch all of the AI, Android reveals

Google Play has a new discovery feature for apps, new ways to acquire users, updates to Play Points, and other enhancements to developer-facing tools.

Google Play preps a new full-screen app discovery feature and adds more developer tools

Soon, Android users will be able to drag and drop AI-generated images directly into their Gmail, Google Messages and other apps.

Gemini on Android becomes more capable and works with Gmail, Messages, YouTube and more

Veo can capture different visual and cinematic styles, including shots of landscapes and timelapses, and make edits and adjustments to already-generated footage.

Google Veo, a serious swing at AI-generated video, debuts at Google I/O 2024

In addition to the body of the emails themselves, the feature will also be able to analyze attachments, like PDFs.

Gemini comes to Gmail to summarize, draft emails, and more

The summaries are created based on Gemini’s analysis of insights from Google Maps’ community of more than 300 million contributors.

Google is bringing Gemini capabilities to Google Maps Platform

Google says that over 100,000 developers already tried the service.

Project IDX, Google’s next-gen IDE, is now in open beta

The system effectively listens for “conversation patterns commonly associated with scams” in-real time. 

Google will use Gemini to detect scams during calls

The standard Gemma models were only available in 2 billion and 7 billion parameter versions, making this quite a step up.

Google announces Gemma 2, a 27B-parameter version of its open model, launching in June

This is a great example of a company using generative AI to open its software to more users.

Google TalkBack will use Gemini to describe images for blind people

Google’s Circle to Search feature will now be able to solve more complex problems across psychics and math word problems. 

Circle to Search is now a better homework helper

People can now search using a video they upload combined with a text query to get an AI overview of the answers they need.

Google experiments with using video to search, thanks to Gemini AI

A search results page based on generative AI as its ranking mechanism will have wide-reaching consequences for online publishers.

Google will soon start using GenAI to organize some search results pages

Google has built a custom Gemini model for search to combine real-time information, Google’s ranking, long context and multimodal features.

Google is adding more AI to its search results

At its Google I/O developer conference, Google on Tuesday announced the next generation of its Tensor Processing Units (TPU) AI chips.

Google’s next-gen TPUs promise a 4.7x performance boost

Google is upgrading Gemini, its AI-powered chatbot, with features aimed at making the experience more ambient and contextually useful.

Google’s Gemini updates: How Project Astra is powering some of I/O’s big reveals

Veo can generate few-seconds-long 1080p video clips given a text prompt.

Google’s image-generating AI gets an upgrade