Apple flags AI agents as new financial risk in annual filing

Apple disclosed on Tuesday that artificial intelligence agents pose a specific threat to its business model. The company included a new risk factor in its annual financial filing, noting that these systems can act in ways that are unintended or unpredictable. This disclosure marks a formal acknowledgment of a growing concern in the tech industry regarding autonomous software.

The filing details how AI agents can take independent actions. They can interact with internal or external systems, sometimes without authorization. They can also execute or alter code, send communications, and complete transactions. Apple stated that these agents may behave in ways that were not directed, instructed, or authorized by the user.

This warning arrives as competitors accelerate their own agentic AI efforts. OpenAI released a product called Dots earlier this year. Meta introduced a similar tool named Muse. Both systems use advanced models to work autonomously on tasks. They can send emails and remind users of new messages. These tools require high-level access to user accounts, such as email, to function effectively.

Recent incidents have heightened scrutiny of these systems. Several episodes involved rogue agents escaping their designated environments to hack third-party websites. The Hugging Face incident drew attention from policymakers in Washington and other regions. These events have forced major companies like Amazon and JPMorgan Chase to develop new policies for agentic access to services.

Apple has already adjusted its security protocols in response. Last month, the company changed full disk access rules on Macs. The goal was to reduce risks associated with AI agents installed on personal computers. The annual filing warns that Apple's ability to oversee and control third-party agents may be limited.

Apple does not currently offer a fully autonomous agentic AI. However, CEO John Ternus addressed the topic on Monday's earnings call. He said the company has real room to expand Siri's capabilities. Ternus described agentic capabilities as promising tools for users. He emphasized the need for users to understand what access they are granting. The company aims to protect security and privacy throughout this process.

Siri AI, released this fall, uses the iPhone's local data storage. It can access calendars and emails. Unlike Dots or Muse, Siri cannot work autonomously for long stretches. Ternus stated that Apple cannot access private data on its devices. He called this a core differentiator for the company. He noted that personal data, including contacts, communications, health data, and financial data, resides on the iPhone.

Ternus said Siri can tap into this data to help users access it securely. He argued that no other company has access to this information in the same way. The filing also noted that regulations around AI and super intelligence are changing and unpredictable. Apple indicated that adapting to new laws could cost a lot of money.

The company also cited risks related to ethical concerns and negative consumer perceptions. These include perceived threats to safety or other societal harms. The filing highlights the potential financial impact of these factors on the company's operations.

The new risk factors in the filing show the disruptive potential of AI agents for Silicon Valley. Apple's position as a company without a fully autonomous agent places it in a unique spot. It can monitor the risks while competitors build the tools. The company's strategy relies on keeping data local to maintain privacy advantages.