INTA 2019: Amazon issues call to arms to fight cyberfraud
Amazon is investing in technology to protect brands on its marketplace, but insists that brands’ cooperation is crucial
The biggest challenge in fighting cyberfraud is the scale of the problem, said Charles Wright, associate general counsel at Amazon, at the INTA Annual Meeting’s May 21 session on “Advances in E-Commerce and Advertising: How Platforms Address Cyberfraud”.
Wright said that users make about five billion changes to Amazon’s website every day, the majority of which are valid actions, which makes bad actors difficult to identify. He said that bot-created accounts are a massive issue due to the data breaches over the last few years that stole millions of individuals’ credentials.
The legal threshold for platforms’ liability for infringing content is knowledge.
“When we have knowledge of a specific infringement, we have an obligation to take it down,” Wright explained. “But we go way beyond that. We’re investing significant resources into proactively keeping fraud out of our stores. This is well in excess of our legal obligations.”
For example, in 2018 alone, more than three billion suspected listings were removed before they could be published, and about one million bad actors’ accounts were stopped before they posted a single listing for sale. Thanks to Amazon’s artificial intelligence (AI) and machine learning (ML) programmes that target cyberfraud, around 100 listings are removed proactively for every one infringement notice that brands submit.
Still, with every improvement in fraud-detection technologies, bad actors have various strategies to stay ahead:
· Diverting fraud, when consumers are linked to other sites to get deals;
· Abusing reviews, when reviews are bought and sold contrary to the nature of honest reviewing;
· Brushing, when people receive packages they never ordered, in order to boost the sellers’ search ranking;
· Falsifying infringement claims, using hacked email addresses; and
· Selling illegal products.
Wright said that there are many more strategies that “I can’t tell you about because we don’t want people to know that we’re onto them”. In terms of counterfeiting specifically, bad actors are wise, Wright said. “They blur your word mark, or get rid of the logo and say ‘product in box bears logo,’ or use 1s instead of Is – it takes a while for us to figure out.”
Amazon’s anti-infringement tools
The main programme that Amazon uses to proactively identify fraudulent postings is Brand Registry. Over 130,000 brands have enrolled in the free programme so far, and Wright said that they submit 99% fewer notices of infringement due to its proactive takedowns based on image-based searching technology.
“We need the inputs from brands so we know who is associated with which marks. When you report a violation, it feeds into our ML loops to help us figure out what we missed and do better next time,” Wright explained.
Transparency is another tool Amazon is using to fight counterfeits. The year-old programme embeds a unique number into each physical product so that every point along the supply chain – even the end consumers – can verify its authenticity. Over 3,000 brands are currently participating.
Amazon’s newest anti-counterfeiting programme is called Project Zero. It’s available by invitation only, and has been live for only three months. It is essentially a self-service takedown tool. Instead of waiting for Amazon to assess the infringement, users have the power to immediately do their own takedowns.
“We’re stepping into this very slowly,” said Wright, “because we’re concerned IP rights owners will try to abuse distribution channels.”
Another project in the works can detect design right infringement. Wright said that he hopes to be able to unveil the new project “later this year”. Until then, infringing images without word logos remain difficult for the company’s AI to detect.
What can brands do in the meantime? Sign up for the Brand Registry programme; it’s free – although it does require a valid registration – and it’s the first step to participating in any of Amazon’s other brand protection initiatives.
“Use the tools we designed,” said Wright. “We’re talking about fighting criminal enterprises attempting to steal money, not paying taxes, spoiling the environment, and avoiding national labour laws. Only through collaboration can we get ahead.”
However, he issued a strict warning that brands should reserve their complaints for instances of actual infringement, instead of unauthorised sales.
“We see this more than we’re happy about,” said Wright. “If it’s simply unauthorised and it’s protected by the first sale doctrine, we’re not going to take it down, and your notice will be rejected. We are keeping track of who sends us these notices,” he concluded.