Image of boat and water.

ASA ruling on Amazon TV ad interacting with AI

Published on 11 April 2018

Was a TV ad which interacted with an Amazon device to place an unwanted order socially irresponsible?

The complaint

A TV ad for the Amazon Echo Dot featured a man's voice instructing: "Alexa, re-order Purina cat food".  The "Alexa" virtual assistant responded: "I've found Purina cat food.  Would you like to buy it?" The complainant argued that the statement: "Alexa, re-order Purina cat food" was socially irresponsible, on the basis that the complainant's Echo Dot had placed an order for cat food after the ad had played.

The response

Amazon confirmed that the complainant's device had a purchase order for Purina Cat Food on the day the ad was seen.  However: 

  • Amazon had technology in place which should prevent its ads from interacting with customer devices.  Advertisements were "marked" so that they did not trigger any responses from Amazon devices when broadcast;

  • if this technology did not work, Amazon had implemented further processes to ensure that an accidental purchase was not made.  Customers had to verbally confirm that they would like to make a purchase for any order to become effective.  If confirmation was not given, the order would be automatically cancelled;

  • in this instance, the order was expressly and immediately cancelled by the customer.  However, had this had not happened, it would have been automatically cancelled due to lack of customer authorization. 

Clearcast stated that they were satisfied that the ad was not socially irresponsible.  They had been assured during the clearance process that there was: (i) a security step in place so that customers would have to verbally confirm an order placed via the Echo, and (ii) technology in place to prevent the advertisement engaging with devices in customers' homes. 

The decision

The ad was investigated under BCAP Rule 1.2 (social responsibility).  The ASA ruled that:

  • Amazon had taken security measures to ensure their ads did not interact with artificial intelligence devices which may overhear them.  In this instance, the technology had failed, causing the device to initiate an order not sought by the customer;

  • however, the additional requirement for the customer to actively confirm their order before a transaction was undertaken meant that it would not be possible for a purchase to be made without the account owner's knowledge, even in cases where technology, implemented to prevent ads and devices interacting, had failed;

  • the ad was not socially irresponsible and did not breach the Code.   

Why is it important?

The decision illustrates that including a requirement for active customer consent provides an additional layer of protection for companies producing interactive devices.  However, the decision also highlights the importance of getting the technology right the first time.

Clearly there is a fine line between a successful interactive AI device and one which is distrusted by consumers for fear that it encroaches on their privacy.  In this instance, even though the purchase was never made, the consumer may have felt uneasy that the device had the potential to make the order in the first place.

Any practical tips? 

Providing the assurance that a device is unable to "spy" on a consumer in their own home is increasingly important in today's climate, as concerns about protection of private information and personal data are frequently driven from the ground-up.

Ensure that any interactive devices include extensive control mechanisms, which not only require the customer to actively consent to any purchases, but also prevent intrusion by the device into the customer's private sphere.