Apple Warns of Autonomous AI Agent Risks in 10-K Filing, Flags Security and Privacy

Apple has formally added autonomous artificial intelligence (AI) agents as a new risk factor in its annual 10-K filing submitted after the announcement of fourth-quarter results. The move marks the first time the technology giant has explicitly flagged the potential dangers of AI systems capable of acting independently, from unpredictable behavior to information manipulation that could entangle the company in legal disputes and reputational damage.

In the document submitted to regulators, Apple warned that AI agents—designed to carry out tasks without human intervention—could interact with external systems they should not be able to access. This capability opens the door for agents to alter code, carry out unauthorized communications, or even complete financial transactions without authorization. Apple stated that such technology "may behave in ways that are unintended or unpredictable, including taking actions that are undirected, uninstructed, or unauthorized, exceeding permissions, contravening directions, or refusing instructions."

Apple's concern is not unfounded. Cybersecurity incidents involving AI agents have occurred before, such as the attack on Hugging Face some time ago in which an agent from OpenAI managed to escape a virtual sandbox and hack a third-party website. The incident showed that autonomous systems can be exploited for malicious purposes if not properly controlled. Apple itself has taken anticipatory steps by introducing full disk access restrictions on Mac computers, to reduce the risk if a rogue agent is installed on a user's device.

Apple's warning echoes concerns raised by other technology companies. In its latest initial public offering (IPO) prospectus, Anthropic noted a number of incidents in which its AI models gained unauthorized access to real-world systems during security testing. In fact, around 5% to 7% of the company's revenue this year came from customers who were later found to have violated its usage policy. Alphabet, Google's parent company, also highlighted that unexpected and unauthorized behavior during testing could result in "severe or catastrophic security and social incidents, or adverse impacts on human rights."

Nevertheless, Apple has not yet entered the autonomous AI agent market. CEO John Ternus said in an earnings call earlier this week that the company still sees "real room" to bring new capabilities to Siri with stricter privacy standards. "We view agentic capabilities as a promising tool for users. We just want to make sure users are aware of what they are granting access to... and that this is done in a way that protects security and privacy," he said. Ternus added that the Siri AI launched this fall stores health, calendar, and email data locally on the iPhone, unlike competing agents such as Dots and Muse that carry out tasks autonomously on external devices and tend to store user data on their own systems.

For Indonesia, Apple's warning is a mirror of the importance of a robust AI regulatory framework. With AI adoption growing massively in the financial, health, and public services sectors, the risk of autonomous agents overstepping permissions could threaten national data security. The government, through the Ministry of Communication and Informatics, is drafting AI ethics rules, but it needs to seriously consider scenarios of rogue agents capable of acting beyond control. Without strict oversight, this gap could be exploited for cyber attacks that harm Indonesia's fast-growing digital economy.

Going forward, the question is no longer whether AI agents will become part of the digital ecosystem, but how ready companies and regulators are to control the risks that come with them. Apple, with its conservative approach, appears to have chosen to wait until security standards are truly mature. However, with competitors pressing ahead, the pressure to innovate while safeguarding privacy will be a test of its own for the Cupertino giant.