Big tech delays AI launches in Europe: A win for privacy or a blow to progress?

The opening event of Apple's annual developer conference this month was one of the most anticipated for the company in a long time. At its center: revealing the company's artificial intelligence (AI) strategy and detailing how modern AI services will be integrated into the company's various products, leveraging personal information stored on devices to answer queries and perform actions that other services, such as OpenAI's ChatGPT and Google’s Gemini, cannot. According to some estimates, if the performance lives up to Apple's promises, it will be a revolution.

But this revolution will not reach, at least not immediately, one major market: the European Union. According to Apple, it will not launch its new AI capabilities this year in Europe due to "regulatory uncertainty" surrounding the Digital Markets Act (DMA) that comes into effect this year. Apple is not alone—Meta and Google have also delayed launching innovative AI products in the Union due to regulatory concerns or limitations.

On the surface, this appears to be a victory for users, competitiveness, the open market, and innovation, showing that sophisticated regulation can prevent potentially harmful actions before they occur. However, the postponement of these services raises the question: does regulation actually harm innovation and produce more harm than good in the long run?

Apple's AI system, Apple Intelligence, will create text and images based on the user's personal context by leveraging information saved on various devices. Despite high expectations, Apple is not expected to launch it in the EU alongside the rest of the world due to the DMA. The law, effective from March 7, creates new regulations for large platforms, including app stores, search engines, browsers, and operating systems. It requires technology companies to limit their use of personal information, limit product exclusivity, allow third-party app installations, and prohibit prioritizing their products over competing ones.In Apple's case, the delay is due to the law's requirement for interoperability, meaning developing software that can work on various operating systems and devices, allowing users to move their information and switch providers easily. Apple claims these requirements "could force us to compromise the integrity of our products in ways that risk user privacy and data security."

Apple's situation is not unique. Recently, Meta delayed the European launch of its AI chatbot, Meta AI, after a privacy regulator asked it to pause plans to train its large language model (LLM) with user posts. Meta had announced it would start training its LLM, Llama, using public posts by Facebook and Instagram users in the EU. Meta explained that the models should be trained on relevant information reflecting the linguistic, geographic, and cultural diversity of European residents. However, the Data Protection Commission of Ireland (DPC), Meta's main privacy regulator in the EU, requested a suspension to examine the possible consequences on residents' privacy.

Similarly, in June 2023, Google announced suspending its GenAI chatbot's planned launch in the EU after the DPC warned that it had not received sufficient information about the product's impact on EU residents' data protection. The chatbot was finally launched a month later. When Google launched the Gemini AI app for Android in May, it reached the EU only a month later.

The tension between technological innovation and regulation often favors the former. Innovative products are launched, and only after years do regulators assess potential damages to competition, privacy, information security, and democratic processes. By then, the companies have profited significantly, and regulators can only impose fines that scarcely affect the companies' bottom lines.

In Apple's and Meta's cases, regulation affects the product even before launch. Apple delayed its AI system to examine regulatory implications after the EU Commission took aggressive actions against it regarding DMA implementation. Recently, the Commission prepared to file an indictment against Apple for not allowing developers to direct users to alternative offers without imposing fees. Apple's announcement signals that the regulation is too burdensome, potentially harming residents' access to advanced services.

In Meta's case, the DPC's proactive step halted the move before it began. Instead of reacting after Meta started training models with user posts, the Commission wants to assess the potential impact on residents' privacy beforehand. They acknowledge Meta's claims that these posts are crucial for the LLM's relevance to European users but want to evaluate the privacy implications.

The reluctance of companies due to new regulation enforcement or proactive regulator actions slows the launch of new technologies. This precautionary approach aims to prevent unforeseen damages from these technologies, but it requires similar practices in other countries to be effective. If only the EU adopts this strategy while other regions like the US and China allow more freedom, European users and businesses may lag in accessing the latest AI products.

There is little reason to believe that the disputes between Apple, Meta, and the EU will be resolved soon. Their product launches in the EU may be delayed for many months, leaving European users behind. Consequently, there is a real concern that while regulation might mitigate competition damage and protect residents' privacy, it could also restrict access to innovative products that could enhance economic activity.

Regulation that staggers the launch of innovative products to contain risks works best if widely adopted. If only one region, even as large as the EU, applies strict regulation, it may result in a loss of innovation and leadership relative to the rest of the world—a potentially too heavy price to pay.