Today, the Romanian Advertising Council, which is the advertising self-regulatory authority in Romania, hosted the conference, Ethics in AI-Created Advertising.  With the laws about the use of AI changing rapidly around the world, the event focused on the question, "What does the responsible use of artificial intelligence in advertising mean today?"

The RAC brought together speakers from Romania and around the world, including Valentin Jucan (Vice President, National Audiovisual Council, Romania), Ludovic Basset (Director General, European Advertising Standards Alliance), Orla Twomey (Chief Executive, Advertising Standards Authority, Ireland), Oana Cociasu (President, Romanian Advertising Council), Dana Dobrescu (Corporate & Government Affairs Manager, Mondelez Romania), and others.  (I was very pleased to be invited to give the U.S. perspective, highlighting the new laws in California and New York, as well as other developments.)

Giving the Romanian government perspective, Valentin Jucan kicked off the conference, highlighting the importance of a strong partnership between government and advertising self-regulation and the need for clear rules about AI use. 

While there were many interesting presentations at the event, EASA's Ludovic Basset provided a very interesting perspective about how advertisers should implement the EU's AI Act.  Article 50 of the AI Act requires, among other things, that users of AI disclose when they use “deep fake” content, with “deep fakes” being roughly defined as AI-generated or manipulated images, audio, or video content that resembles existing persons, objects, places, entities, or events, that falsely appear to be authentic.

Basset talked about the important question that global and European advertisers are thinking about right now, which is, does the EU's AI Act require that every ad that includes AI-generated or modified content be labeled?  Noting that “our goal should be smarter transparency,” he said that the AI Act doesn't require disclosure in every situation, and that advertisers need to know where to draw the line. 

Basset said that realism is not the only test for when disclosure is required, but that instead advertisers need to look at the context in which AI-generated or modified content is used in order to assess what the impact is on consumers.  He said that advertisers need to ask three key questions:

• How substantially was AI used?
• How plausible or realistic is the AI content?
• How significant is the AI content for the consumer? 

Basset said that, by asking these questions, advertisers can better assess whether disclosure is actually required.  Putting this into practice, he explained that when AI is used to alter the background of an image, labeling is not automatically required; rather, advertisers need to assess the whole ad to consider the context.  Similarly, the AI Act doesn't ordinarily require disclosure when AI is used to make minor, technical edits to the advertising.  On the other hand, the AI Act would require disclosure when  AI is used in a material manner, such as when a celebrity deepfake is used.  Finally, he emphasized that truth in advertising principles still apply.  Noting that even if an advertisement is labeled as AI-generated, advertisers still need to consider whether the advertising is misleading. He explained, “a label cannot cure a misleading claim.” 

Basset also discussed the importance of paying attention to national implementation of Article 50 as well as the guidance being issued from local advertising self-regulatory organizations. He encouraged advertisers to pay attention to nuance and to aim for “meaningful transparency.”