Recommerce Reads – Innovating resale: AI considerations for online second-hand retail

Recent advances in artificial intelligence (“AI”), including in relation to generative AI tools, mean that AI-related stories have become front page news.

AI presents a number of benefits and opportunities for businesses, including those operating in the recommerce space. However, before using AI, it is important that businesses are aware of and take steps to mitigate the potential risks associated with AI. In addition, businesses (including UK businesses) using AI in the EU need to prepare for compliance with the EU AI Act, which is due to start applying (in part) from 2 February 2025. We expect AI-related legislation or regulation to follow in other jurisdictions, including the UK (where the new Labour government intends to introduce AI legislation in respect of the most powerful AI models), and so businesses will need to ensure that they are compliant with any applicable AI requirements.

In this article, we discuss some of the considerations relevant to current or future uses of AI in the online second-hand retail market.

1. Conversational recommerce

For recommerce businesses looking to improve their customer service solutions, AI-powered chatbots and voice assistants can answer queries, help customers find the items they are looking for, guide customers through the purchase process and provide enhanced customer support.

Under the EU AI Act, transparency obligations will apply in respect of chatbots intended to interact directly with individual customers. These chatbots will need to be developed in a way that customers are informed they are interacting with an AI-powered chatbot, unless this would be obvious in the circumstances.

Where a chatbot is used to detect customer’s emotions so that it can escalate to a human customer support agent as appropriate (for example, if the chatbot detects that the customer is angry), the chatbot may also be subject to the EU AI Act’s obligations on high-risk AI systems.

In 2022, a chatbot on the AirCanada website informed a customer that he could apply for a discount after booking a flight. The customer booked the full-fare flight in reliance on this information, only to discover subsequently that it was incorrect. AirCanada argued initially that the chatbot was responsible for its own actions, but the court found that AirCanada was liable and had to pay damages to the customer. Although this case was brought in Canada, it nonetheless highlights the need for businesses to take steps to improve the reliability and accuracy of any chatbots and to put in place sufficient guardrails when using chatbots. Given the PR fallout caused by AirCanada’s initial reaction, this suggests that one way for businesses to mitigate the risk of reputational damage in these scenarios will be to take responsibility for their chatbots from the outset.

2. AI-generated content

The online nature of the recommerce space means that customers cannot physically inspect the item they are looking to buy. Online retailers may therefore be highly motivated to provide a detailed digital representation of the products they are offering, to improve customer confidence and increase the number of sales. The same applies for platforms facilitating second-hand peer-to-peer sales. Recommerce retailers and platforms can innovate their sales environment by leveraging AI tools to generate product listings, including automating product descriptions and images.

Content generated by AI tools may fall under the transparency obligations in the EU AI Act, which require artificially generated or manipulated content to be labelled or disclosed accordingly.

Where using AI tools to generate content, businesses also need to be aware of potential copyright issues. In the absence of guidance from the courts, it’s not clear how questions of subsistence, authorship, ownership and infringement will be answered in the AI context. In connection with this, businesses should check if the terms and conditions of any third-party AI tools make clear whether the business (as the user) owns any copyright in the AI-generated content and/or whether the relevant third party accepts any responsibility for third-party copyright infringement claims.

Generative AI tools can sometimes produce inaccurate content (often called “hallucinations”) or content that perpetuates biases, including in relation to race and gender. There have also been examples of chatbots being ‘tricked’ into swearing or providing offensive responses. To avoid reputational damage in this respect, businesses need to ensure that any AI-generated content has been reviewed by a human before it is published or posted so that any incorrect, biased or offensive content can be identified and either corrected or removed as appropriate.

3. Virtual stylists

One of the challenges for an online retailer, including in the second-hand fashion market, is the lack of a personalised shopping experience for each customer. Where businesses are looking to provide personalised styling advice, AI-powered virtual styling tools could be used to build, track and manage customer styling profiles, providing personalised recommendations based on their preferences and, in some cases, based on analysing customer photos. For example, there are virtual styling tools that use a selfie uploaded by a customer to analyse that customer’s facial features (including skin tone) and then, based on that analysis, suggest items of clothing for that particular customer.

Businesses will need to take care when using AI tools to analyse an individual’s face, especially if analysing an individual’s skin tone or colour, as this may, under the EU AI Act, fall within the list of banned AI practices or otherwise be subject to obligations on high-risk AI systems. In the absence of guidelines on the implementation of the EU AI Act, it’s not entirely clear how this type of tool would be regulated (or whether it would fall within one of the exceptions available under the EU AI Act), but this will need to be assessed carefully on a case-by-case basis.

4. Finding the right fit

The top reason given for returning clothes purchased online is that they didn’t fit properly. One of the most common uses of AI-powered tools in online retail is to help customers to identify the right size for them. This includes tools using computer vision to analyse customer photos to determine that customer’s measurements and recommend the right size for a particular brand. This could be particularly useful in the recommerce sector, where the range of sizing across a wide selection of second-hand clothing will vary across brands and decades.

As with some of the other uses mentioned in this article, using AI to analyse photos of a customer gives rise to various privacy or data protection considerations. Recommerce operators intending to deploy this type of tool will need to ensure they are compliant with applicable data protection law and, where using a third party’s tool, should undertake appropriate due diligence in relation to that third party to ensure there are appropriate safeguards around customer personal data.

Next steps

AI has the potential to transform the online second-hand retail sector, offering solutions to many of the challenges created by the move away from ‘brick-and-mortar’ stores and the variety of product selection. Recommerce businesses should consider carefully how to adopt AI in a manner that complies with applicable laws and ensures there are appropriate contractual protections when procuring third-party AI services or solutions. While the potential fines payable for non-compliance with applicable law (such as the EU AI Act or the GDPR) demonstrate the importance of getting it right, arguably, the real incentive for using AI in a compliant manner is that it can improve customer trust and confidence, resulting in increased sales of pre-loved goods.

For more articles on topics affecting recommerce businesses, including greenwashing, and sponsorships and endorsements, please visit our Recommerce webpage. If you would like to discuss AI considerations in the recommerce sector further, please do not hesitate to contact one of our specialists.