Government & Policy

Amazon faces more EU scrutiny over recommender algorithms and ads transparency

Comment

Amazon Fulfilment Center In Sosnowiec
Image Credits: Beata Zawrzel/NurPhoto / Getty Images

In its latest step targeting a major marketplace, the European Commission sent Amazon another request for information (RFI) Friday in relation to its compliance under the bloc’s rulebook for digital services.

The development highlights areas where EU enforcers are dialing up their scrutiny of the e-commerce giant, with the bloc asking for more info about Amazon’s recommender systems, ads transparency provisions and risk assessment measures.

An earlier Commission RFI to Amazon, last November, focused on risk assessments and mitigations around the dissemination of illegal products; and the protection of fundamental rights, including in relation to its recommender systems. A Commission spokesperson confirmed the e-commerce giant has received three RFIs in all — following a January ask for more info on how it’s providing data access for researchers.

The EU’s Digital Services Act (DSA) puts requirements on platforms and services to abide by a series of governance standards, including in areas like content moderation. In the case of online marketplaces the law also requires they implement measures to enable them to take action to tackle risks around the sale of illegal goods. Larger marketplaces, such as Amazon, have an additional layer of algorithmic transparency and accountability obligations under the regime — and this is where the Commission RFIs are focused.

The additional rules have applied on Amazon since the end of August last year, following its designation by the EU as a very large online platform (VLOP) in April 2023. It’s the Commission’s job to enforce these extra obligations on VLOPs.

While it remains to be seen if the latest Commission RFI to Amazon will lead to a formal investigation of its DSA compliance the stakes remain high for the e-commerce giant. Any confirmed violations could get very costly as penalties for breaching the pan-EU law can reach up to 6% of global annual turnover. (NB: The company’s full-year revenue for 2023 was $574.8 billion, meaning — on paper at least — its regulatory risk runs into double-figure billions.)

Detailing its action in a press release, the Commission said it has sent Amazon an RFI related to measures it has taken to comply with DSA rules related to the transparency of recommender systems and their parameters. It also said it’s asking for more info about Amazon’s provisions for maintaining an ad repository — another legally mandated transparency step for larger platforms.

The Commission also said it wants more detail about Amazon’s risk assessment report. The DSA requires VLOPs to both proactively assess systemic risks that might arise on their platforms and take steps to mitigate issues. Platforms also need to document their compliance process.

“In particular, Amazon is asked to provide detailed information on its compliance with the provisions concerning transparency of the recommender systems, the input factors, features, signals, information and metadata applied for such systems and options offered to users to opt out of being profiled for the recommender systems,” the EU wrote. “The company also has to provide more information on the design, development, deployment, testing and maintenance of the online interface of Amazon Store’s Ad Library and supporting documents regarding its risk assessment report.”

The EU has given Amazon until July 26 to provide the requested info. After that, any next steps will depend on its assessment of its response. But failure to respond satisfactorily to an RFI could itself trigger a sanction.

Last year the EU named online marketplaces as one of a handful of priority issues for its enforcement of the DSA’s rules for VLOPs. And it has looked attentive to the area.

Late last month it sent separate RFIs to rival marketplace VLOPs, Shein and Temu — soon after designating the pair. Although, in their case, the Commission’s RFIs also raised concern about illegal goods risks and manipulative design (including as a potential child safety risk), as well as asking them for more information about the operation of their own recommender systems.

Why so much interest here? Algorithmic sorting has the power to influence platform users’ whole experience by determining the content and/or products they see.

In a nutshell, the EU wants the DSA to crack open such blackbox AI systems to ensure that platforms’ commercial agendas — to grab users’ attention and/or drive more sales — aren’t the only thing programming these automated decisions. It therefore wants the DSA to act as a shield against the risks of AI-driven societal harms, such as platforms pushing content that’s harmful for people’s mental health or recommending shoppers buy dangerous products. But achieving that goal will require enforcement.

Amazon, meanwhile, is unhappy about the EU regime. Last year it challenged its DSA designation as a VLOP. And last fall it won an interim stay on one element of VLOPs’ DSA compliance — namely the requirement to publish an ads library. However, in March, the EU General Court reversed the earlier decision, overturning the partial suspension.

“Following its designation as a Very Large Online Platform and the Court’s decision to reject Amazon’s request to suspend the obligation to make its advertisement repository publicly available, Amazon is required to comply with the full set of DSA obligations,” the Commission wrote today. “This includes diligently identifying and assessing all systemic risks relevant to its service, providing an option in their recommender systems that is not based on user profiling, and have an advertisement repository publicly available.”

Given Amazon has spent money on lawyers to try to argue why it shouldn’t have to comply with the DSA ads library element — and the subsequent overturning of the stay — it’s not too surprising this is one of the areas where the Commission is seeking more information now.

The EU was contacted with questions. A Commission spokesperson confirmed the first RFI to Amazon, from November 2023, had “a strong focus on the dissemination of illegal products and the protection of fundamental rights online”, as well as asking questions about its recommender systems.

A second RFI, in January 2024, focused on measures Amazon has taken to comply with data access for eligible researchers, per the spokesperson. They said the latest RFI is strongly focused on measures taken to meet DSA obligations related to the transparency of recommender systems and their parameters, as well as to the provisions on maintaining an ad repository.

“These are actually different areas we are looking into,” the spokesperson added. “You are however right to say that today’s RFI also follows the Court’s decision to reject Amazon’s request to suspend the obligation to make its advertisement repository publicly available.”

We also reached out to Amazon for a response to the Commission’s RFI.

A company spokesperson emailed TechCrunch this statement: “We are reviewing this request and working closely with the European Commission. Amazon shares the goal of the European Commission to create a safe, predictable and trusted shopping environment. We think this is important for all participants in the retail industry, and we invest significantly in protecting our store from bad actors, illegal content, and in creating a trustworthy shopping experience. We have built on this strong foundation for DSA compliance.”

This report was updated with responses from the Commission

More TechCrunch

Prior to the ban, Trump’s team used his channel to broadcast some of his campaigns. With the ban now lifted, his channel can resume doing so.

Twitch reinstates Trump’s account ahead of the 2024 presidential election

Featured Article

Faulty CrowdStrike update causes major global IT outage, taking out banks, airlines and businesses globally

Security giant CrowdStrike said the outage was not caused by a cyberattack, as businesses anticipate widespread disruption.

Faulty CrowdStrike update causes major global IT outage, taking out banks, airlines and businesses globally

The CrowdStrike outage that hit early Friday morning and knocked out computers running Microsoft Windows has grounded flights globally. Major U.S. airlines including United Airlines, American Airlines and Delta Air…

CrowdStrike outage: How your plane, train and automobile travel may be affected

This week, Google is in discussions to pay $23 billion for cloud security startup Wiz, SoftBank acquires Graphcore, and more.

M&A activity heats up with Wiz, Graphcore, etc.

CrowdStrike competes with a number of vendors, including SentinelOne and Palo Alto Networks but also Microsoft, Trellix, Trend Micro and Sophos, in the endpoint security market.

CrowdStrike’s rivals stand to benefit from its update fail debacle

The IT outage may have an unexpected effect on the climate: clearer skies and maybe lower temperatures this evening

CrowdStrike chaos leads to grounded aircraft — and maybe an unusual weather effect

There’s a man in Florida right now who wants to propose to his girlfriend while they’re on a beach vacation. He couldn’t get the engagement ring before he flew down…

The CrowdStrike outage is a plot point in a rom-com 

Here’s everything you need to know so far about the global outages caused by CrowdStrike’s buggy software update.

What we know about CrowdStrike’s update fail that’s causing global outages and travel chaos

This serves as an example for how easy it is to spread inaccurate information online during a time of immense global confusion and panic.

From the Sphere to false cyberattack claims, misinformation runs rampant amid CrowdStrike outage

Today is the final chance to save up to $800 on TechCrunch Disrupt 2024 tickets. Disrupt Deal Days event will end tonight at 11:59 p.m. PT. Don’t miss out on…

Last chance today: Secure major savings for TechCrunch Disrupt 2024!

Indian fintech Paytm’s struggles won’t seem to end. The company on Friday reported that its revenue declined by 36% and its loss more than doubled in the first quarter as…

Paytm loss widens and revenue shrinks as it grapples with regulatory clampdown

J. Michael Cline, the co-founder of Fandango and multiple other startups over his multi-decade career, died after falling from a Manhattan hotel, New York’s Deputy Commissioner of Public Information tells…

Fandango founder dies in fall from Manhattan skyscraper

Venture capital giant a16z fixed a security vulnerability in one of the firm’s websites after being warned by a security researcher.

Researcher finds flaw in a16z website that exposed some company data

Apple on Thursday announced its upcoming lineup of immersive video content for the Vision Pro. The list includes behind-the-scenes footage of the 2024 NBA All-Star Weekend, an immersive performance by…

Apple Vision Pro debuts immersive content featuring NBA players, The Weeknd and more

Biden centering Musk in his campaign is a notable escalation, considering he spent most of his presidency seemingly pretending the billionaire didn’t exist.

Elon Musk is now a villain in Joe Biden’s presidential campaign

Waymo would need a ground transportation permit to operate at SFO, which has yet to be approved.

Waymo wants to bring robotaxis to SFO, emails show

When Tade Oyerinde first set out to fundraise for his startup, Campus, a fully accredited online community college, it was incredibly difficult. VCs have backed for-profit education companies in the…

Why it made sense for an online community college to raise venture capital

Canadian private equity firm PartnerOne paid $28.2 million for HeadSpin, a mobile app testing startup whose founder was sentenced for fraud earlier this year, according to documents viewed by TechCrunch.…

PE firm PartnerOne paid $28M for HeadSpin, a fraction of its $1.1B valuation set by ICONIQ and Dell Technologies Capital

Meta has suspended the use of its AI assistant after Brazil’s National Data Protection Authority (ANPD) banned the company from training its AI models on personal data from Brazilians. The…

Meta puts a halt to training its generative AI tools in Brazil 

ChatGPT, OpenAI’s text-generating AI chatbot, has taken the world by storm since its launch in November 2022. What started as a tool to hyper-charge productivity through writing essays and code…

ChatGPT: Everything you need to know about the AI-powered chatbot

The Mumbai-based firm said one of its multisig wallets had suffered a security breach, and it was temporarily pausing all withdrawals from the platform.

WazirX halts withdrawals after losing $230 million, nearly half its reserves

This week’s TechCrunch Mobility looks at Fisker scoring a win, an AV startup rebooting in Texas, why Elon is pushing the Tesla robotaxi reveal and more.

Fisker scores a win, an AV startup reboots in Texas, and why Elon pushed the Tesla robotaxi reveal

Apple Intelligence was designed to leverage things that generative AI already does well, like text and image generation, to improve upon existing features.

What is Apple Intelligence, when is it coming and who will get it?

The European Union’s president, Ursula von der Leyen, was confirmed in the role for another five years Thursday after parliamentarians voted overwhelmingly to re-elect her. The scale of her support…

The EU just re-elected its president for another five years — here’s what that means for tech

Olivia DeRamus is flipping the script: What if scrolling through social media didn’t make us miserable? What if, especially for women, social media could actually make us feel more supported?…

Communia bets social media can be good for you

TikTok is partnering with the music distribution service DistroKid to fast-track the creation of artist accounts for members. The ByteDance-owned short video platform introduced an Artist Account feature last year…

TikTok fast-tracks artist account creation for DistroKid members

Ford is still pushing forward on electrification, notably by increasing hybrid options.

Ford’s EV plans are in flux once again as it invests $3B into its biggest trucks

OpenAI introduced GPT-4o mini on Thursday, its latest small AI model. The company says GPT-4o mini, which is cheaper and faster than OpenAI’s current cutting-edge AI models, is being released…

OpenAI unveils GPT-4o mini, a smaller and cheaper AI model

Featured Article

USPS shared customer postal addresses with Meta, LinkedIn and Snap

The U.S. Postal Service confirmed it took action to “remediate” the data sharing following a TechCrunch investigation.

USPS shared customer postal addresses with Meta, LinkedIn and Snap

The automotive industry is in the midst of dramatic technological change as companies seek out new ways to make money beyond building and selling gas-powered cars. And GM CEO and…

GM CEO Mary Barra is coming to TechCrunch Disrupt 2024