Welcome to the Era of the Universal AI Assistant

For years, the expense of utilizing “free” services from Google, Facebook, Microsoft, and other tech giants has involved relinquishing your data. Sharing your life online and taking advantage of free technology offers convenience, but it also places personal information in the hands of large corporations that often aim to profit from it. Now, the emerging wave of generative AI systems is expected to demand even greater access to your data.

In the last two years, generative AI tools—like OpenAI’s ChatGPT and Google’s Gemini—have evolved beyond the simplistic, text-only chatbots initially introduced by these companies. Instead, Big AI is increasingly focusing on the development and promotion of agents and “assistants” that claim they can perform tasks and take actions on your behalf. The catch? To maximize their potential, you’ll need to provide them access to your systems and data. While much of the early debate surrounding large language models (LLMs) revolved around the blatant copying of copyrighted material online, granting AI agents access to your personal data is likely to introduce a new wave of challenges.

“AI agents, in order to function fully and access applications, often need to reach the operating system level of the device running them,” explains Harry Farmer, a senior researcher at the Ada Lovelace Institute, whose studies on AI assistants indicate they may pose a “profound threat” to cybersecurity and privacy. For tailored experiences with chatbots or assistants, Farmer notes that there are often data trade-offs involved. “All those functionalities require substantial information about you,” he states.

While there isn’t a concrete definition of what an AI agent is, they can be best understood as a generative AI system or LLM that has been granted a degree of autonomy. Currently, agents or assistants, including AI web browsers, can take control of your device to browse the internet on your behalf, book flights, conduct research, or add items to shopping carts—some capable of executing tasks comprising multiple steps.

Although present-day AI agents may be prone to glitches and frequently struggle to complete assigned tasks, tech companies are betting these systems will radically transform millions of jobs as they enhance their capabilities. A crucial aspect of their effectiveness likely stems from data access. Therefore, if you want a system to manage your schedule and tasks, it will require access to your calendar, messages, emails, and more.

Some more sophisticated AI products and features hint at the level of access agents and systems could possess. Specific agents developed for business purposes can analyze code, emails, databases, Slack messages, files in Google Drive, and more. Microsoft’s contentious Recall product captures screenshots of your desktop every few seconds, enabling you to search everything you’ve done on your device. Tinder has introduced an AI feature that can sift through photos on your phone to “better understand” users’ “interests and personality.”

Carissa Véliz, an author and associate professor at the University of Oxford, points out that, in most cases, consumers lack a reliable way to verify if AI or tech companies are managing their data as they assert. “These companies are very lax with data,” Véliz remarks. “They have demonstrated a lack of respect for privacy.”