How Italy’s ChatGPT ban is disrupting Wunderman Thompson’s local office – AdAge.com

Italy’s ban of ChatGPT, the popular generative AI chatbot, has incited concern for many in the marketing world who have been experimenting with the technology. And for at least one Italy-based ad agency, the embargo has resulted in the total suspension of the tool. 

Italy became the first Western country to ban online access to ChatGPT last week, and the nation’s data protection authority explained the decision as stemming from concerns over digital privacy. The governing body stated the chatbot has no legal right to obtain personal data from users for the purpose of further training.

In the marketing industry, ad agencies, including those in Italy, are actively experimenting with ChatGPT. But with the sudden embargo, these local firms have had to cease all activity. One such shop is Wunderman Thompson Italy, which has not only been forced to suspend its study of the tool but is also facing disruptions to new patterns of conducting business.

While the ban is localized, marketers everywhere may face ramifications. Italy is the first government to draw a line in the sand, but it is expected to not be the last, as regulating bodies across the world discuss how to reconcile the advancing AI space with existing legislation. The case of Italy exemplifies an early application of one such effort.

Also read: Inside agencies’ AI practices

Wunderman Thompson is a global agency, with an office based out of Milan. For this cohort, the Italian government’s ban came as a big surprise, said Marco Bandini, chief technology officer at Wunderman Thompson Italy.

“It’s not normal to see Italy on the side of China and Russia,” he said. The only other countries to ban ChatGPT are those with high levels of internet censorship, such as China, Russia, Iran and North Korea.

Since ChatGPT rolled out to the public at the end of last year, Wunderman Thompson Italy has encouraged its employees to engage with the tool. The agency offered a course in prompting, or instructing the bot to complete a task, to every department. Wunderman’s goal was to get everyone in the company well-versed with the power of generative AI, said Bandini.

Even before ChatGPT’s launch in November, the agency had used the bot’s foundational technology in work with clients. A 2021 project with Campari, for example, hinged on leveraging OpenAI’s GPT-2 model to develop a script in the style of iconic Italian filmmaker Frederico Fellini.

Wunderman’s embrace of generative AI bred new employee behaviors, such as relying on ChatGPT for data analysis, as well as using the bot as a stand-in colleague with whom to brainstorm ideas. 

But now that ChatGPT has been blocked, these patterns have been completely disrupted, Bandini said. More broadly, the suspension has alienated the firm as colleagues around the globe are still able to use the tool freely.

Wunderman remains permitted to use ChatGPT’s application programming interface (API), which does not collect user data. But this model is less capable than the online version, and because it doesn’t process user data, its ability to learn is limited.

Bandini said the agency continues to use other generative AI tools, such as image generators DALL-E—also created by OpenAI, the company behind ChatGPT—and Midjourney. Image-to-text bots have been used by marketers to bring ideas to life. Yet they are not as well-rounded as ChatGPT, which is able to perform all kinds of clerical work in addition to creativity.

Related: AI and Midjourney lessons from a creative director

There are workarounds for Italian marketers who refuse to take no for an answer, said Bandini. PizzaGPT is an accessible platform created by a pseudonymous individual that is built on OpenAI’s GPT 3.5 technology, although its performance has apparently been patchy. The model gets its name from its request for donations equivalent to the price of a pizza.

Virtual private networks (VPNs) are another alternative, Bandini said, which allow users to hide their IP addresses and thus circumvent restrictions on their location. VPNs are commonly used for privacy and security purposes, and are entirely legal in Italy.

Related: AI copyright ruling could pose issues for agencies

What happens next?

OpenAI this week met with Italy’s data protection authority after OpenAI sent a letter expressing a willingness to find a solution, Reuters reported. If the ban lifts, Danilo Maurizio, head of data at Wunderman Thompson Italy, anticipates it will come with new limitations, such as restrictions on age and other mechanisms of gating access.

How, or whether, these concessions will affect Italian ad agencies remains unclear. There is also the potential for Italy’s decision to pose ramifications for other Western countries. Since the EU’s privacy framework, the General Data Protection Regulation (GDPR), does not adequately address generative AI, other nations within the union may follow Italy’s move, said Maurizio.

“The [Italian] government said, ‘We must do something’—this is something,” he told Ad Age. Although Maurizio expressed optimism that the ban could impel more responsible adoption of AI tools, Maurizio stopped short of supporting its outright prohibition. He fears the emergence of an unregulated secondary market for large language models (LLMs) like ChatGPT if the bot is disallowed for too long.

Thierry Breton, the EU’s commissioner for internal market, has already made clear that the group is discussing new AI-focused rules. Britain, which is no longer a part of the EU, has asked regulators to apply existing policies to the AI space.

Talks of limiting AI’s reach have even cropped up in the U.S. Last month, a group of over 1,000 AI experts, including Elon Musk, issued an open letter that demanded a pause to the development of models more advanced than OpenAI’s GPT-4. While not the same as Italy’s action, the letter is a sign that limitations on the use of generative AI may already be in the works.

Adblock test (Why?)