A workshop on deepfakes took place during the MARQUES Spring Team Meeting in Barcelona today. Around 200 MARQUES Council and Team members are attending the Spring Meeting this week.

The speakers at the workshop were Julia Holden (Trevisan & Cuonzo, Italy), Thomas Hertl (Arnecke Sibeth Dabelstein, Germany) and Maiken Toftgaard (Poul Schmith/Kammeradvokaten, Denmark) of the MARQUES Copyright Team and Graeme Murray (Marks & Clerk, UK) and Howard Shire (Troutman Pepper Hamilton Sanders, US) of the MARQUES Education Team. The workshop was moderated by Sandra Müller (Squire Patton Blogs, Germany) of the Education Team.

Introducing the workshop, Sandra said deepfakes are easy to create and hard to protect. They are one of the biggest threats to society today and they also impact brands. She said the workshop would address what tools can IP practitioners use to address this threat?

What are deepfakes?

Graeme kicked off the workshop by defining disinformation and deepfakes. He said the latter encompasses brand manipulation, political manipulation, entertainment manipulation and fraudulent deepfakes.

He also shared examples of deepfakes, including those featuring Bruce Willis, the Pope wearing a puffer jacket and Elon Musk kissing Giorgia Meloni, illustrating both how realistic and how damaging they can be.

Legal tools

The 2022 EU Code of Practice on Disinformation has at least 44 signatories (including Google, Meta, X and TikTok) who have committed to tackling the spread of disinformation online. There are efforts to encourage more signatories to join.

The Commission sets goals but they are not legally binding on signatories. Julia described the Code as a “toolkit to fight disinformation” that is “reasonably effective”.

For example, under the Code very large online platforms report every six months on what they are doing on implementation.

By contrast the EU Digital Services Act is legally binding for online platforms and intermediaries regarding illegal content, imposing stricter regulations on the largest players for the dissemination of harmful data. However, the Act does not define disinformation.

The EU AI Act can also play a role in regulating deepfakes. Maiken said that while it does not ban deepfakes it imposes certain disclosure obligations.

The Act takes a risk-based approach to AI tools: most deepfakes would likely be classified as low risk unless they are political in which case there would be additional requirements.

The panel also discussed country-specific regulations, including criminal and administrative provisions, in Italy, Germany, Denmark, the UK and the US – and in various US states.

Tips and strategies for brand owners

During the Q&A part of the workshop, the panelists discussed challenges including who can take action, the risks for brands of attacking deepfakes, how to act in a timely way and how to prove damage.

Advice included: crawling the internet; working with social media operators; take a case-by-case approach; have a deepfake strategy in place including identifying the source; and work with marketing to get consents.

The Spring Meeting continues with Team meetings today, followed by a dinner and the annual Kay Uwe Jonas Lecture this evening and Team presentations tomorrow.