Apple’s annual report for fiscal 2026, filed with the SEC on 3 November, warns that AI agents and similar technologies “may behave in unintended or unpredictable ways, including by taking actions that were not directed, instructed or authorized, exceeding permissions, defying directions or resisting instructions”. The sentence sits in Apple’s risk factor on artificial intelligence, which it first added in its 10-Q of 1 May. That factor has grown from 199 to 628 words, by our count.
Across the whole risk-factor section, the phrase “artificial intelligence” now appears 45 times, against six in the fiscal 2025 10-K and three in the fiscal 2024 10-K, our count. With this filing Apple becomes the seventh of the 12 companies whose risk factors we read for agent wording to warn of unintended or unauthorized actions by AI systems.
Business Pill · Least Privilege
An evergreen Business Engineer explainer on least privilege: giving each user or system only the access it needs. It explains the general idea and does not discuss Apple.
The key insight: By our count, Apple’s risk factors name artificial intelligence 45 times, up from six a year earlier, and its AI risk factor has grown from 199 to 628 words since it first appeared in May, now including a paragraph on AI agents that may act without authorisation.
The May version of the AI risk factor listed the risks Apple’s AI efforts may give rise to, from competition to the protection of minors. The 10-K adds four items to that list. Here is the new run of the list, in which only the second item was there in May: “unauthorized access to or inadvertent releases of confidential information or other unexpected access to or use of user information; exposing users to harmful, inaccurate or other negative content or experiences; unintended or unauthorized actions by autonomous or agentic systems; misuse of artificial intelligence by employees and other personnel; changing public sentiment”.
It then adds a paragraph on agents. It opens: “In particular, artificial intelligence agents and other technologies can take actions autonomously, including interacting with internal or external systems that in some cases they may not be authorized to access, executing or altering code, sending communications, and completing transactions”.
The paragraph says these technologies, whether built by Apple or by third parties, may act without instruction or exceed permissions. It continues: “In addition, such technologies may conceal actions, hide, withhold or manipulate information, or generate inaccurate or deceptive outputs”. It also says “Some consequences, such as deleting data or completing transactions, may be difficult or impossible to reverse”, and: “Moreover, the Companyโs ability to oversee and control third-party artificial intelligence agents and other technologies may be limited”.
None of these passages appears in the fiscal 2025 10-K or in the three 10-Qs Apple filed in 2026, by our check.
Left: mentions of “artificial intelligence” in Item 1A of Apple’s 10-Ks for fiscal 2024, 2025 and 2026. Right: the length of Apple’s AI risk factor when it first appeared, in the 10-Q filed on 1 May 2026, and in the fiscal 2026 10-K. Counts are our count from SEC EDGAR text.
Agents and the Privacy Pitch
A second new passage sits in the risk factor on competition. It says “the introduction of emerging technologies, such as artificial intelligence agents that require access to personal data, creates privacy and data security concerns that could erode customer trust, diminish the perceived value of the Companyโs privacy protections, and undermine its competitive advantage”.
The same factor now says: “Failure to successfully develop, deploy or differentiate artificial intelligence capabilities, or to do so as quickly or cost-effectively as competitors, could result in loss of market share, erosion of competitive positioning, and inability to attract or retain customers and talent”. Chief executive John Ternus cited privacy nine times on his first earnings call, on 2 November, and three analysts asked about agents.
A First Mention of ‘Super Intelligence’
The 10-K uses the term “super intelligence” four times in its risk factors, our count. The fiscal 2025 10-K and the three 10-Qs Apple filed in 2026 do not use it.
One mention is in the trade risk, where export controls can cover “technologies, including artificial intelligence and super intelligence technologies, or data”. Another opens the regulatory part of the AI factor: “The divergent and rapidly evolving nature of legal and regulatory frameworks around artificial intelligence and super intelligence and increased scrutiny from lawmakers and regulators create significant uncertainty”. A third covers “ethical concerns and negative consumer perceptions related to automation, artificial intelligence and super intelligence”.
Compute, Data Centres and Memory
The investment risk factor now gives AI as its example: “For example, the Company has made significant investments in artificial intelligence technologies, infrastructure and personnel, including commitments to secure data center capacity, components, equipment and energy resources”. It adds that “the usage of artificial intelligence may not enhance the Companyโs products or services or be beneficial or profitable for its business”.
The components and compute factor, which we covered when it first named compute, is now 652 words, against 542 in the June 10-Q and 279 in the fiscal 2025 10-K, by our count. The 10-K adds storage to compute, and says: “There are limited third-party providers that can reliably meet such demand”.
It also lists what can slow Apple’s and its providers’ data centre builds, including “the availability of power, water and land, labor shortages, public sentiment”. In June, Apple said it expected component supply constraints and cost increases to intensify. The 10-K narrows that: “The Company expects these trends to intensify for memory and storage components”. Apple’s supplier obligations are the subject of a separate piece.
What the Risk Factors Say
The risk-factor section grew to 12,212 words from 9,686 in the fiscal 2025 10-K, 26% longer, by our count. Legal, antitrust and tax risks remain the largest theme, at 29.3% of the words by our classification. AI regulation and policy now takes 5.3%, a theme that had no factor of its own a year earlier.
Several AI passages are not new in this filing. The cyberattack wording on phishing and deepfakes was already in the March-quarter 10-Q, and the sentence warning that Siri AI may not launch in some jurisdictions because of interoperability rules was already in the June 10-Q.
Seven of 12
Before this filing, six of the 12 companies we read carried agent-action wording: Palantir, Alphabet, Microsoft, Anthropic, Meta and Salesforce. Apple’s June 10-Q did not; Apple’s 10-Qs list only changes to the annual report. Its 10-K now does, which makes seven, by our count.
The Structural Read
As we read it, the agent paragraph covers agents Apple builds and agents others build, and adds that Apple’s ability to oversee third-party agents may be limited.
As we read the competition factor, Apple now ties AI agents that need personal data to the perceived value of its privacy protections, the theme its chief executive stressed on the 2 November call.
Apple Inc., Form 10-K for fiscal 2026, risk factors
“Some consequences, such as deleting data or completing transactions, may be difficult or impossible to reverse.”
Three Implications
FOR THE AGENT COUNTApple is the seventh of the 12 companies in our review whose risk factors warn of unintended or unauthorized actions by AI systems, by our count.
FOR THE PRIVACY PITCH The 10-K says AI agents that require access to personal data could diminish the perceived value of Apple’s privacy protections.
FOR COMPUTE AND MEMORY The 10-K says there are limited third-party providers that can reliably meet compute demand and that Apple expects supply constraints and rising costs to intensify for memory and storage.
The framework puts it this way: “The governance layer sits at the intersection of capability, regulation, national security, access control, and geopolitical alignment.”
As we read it, Apple’s new agent paragraph describes the access-control part of that layer from the side of a device maker whose users may run agents built by others. The framework does not discuss Apple.
This piece is based on Apple’s fiscal 2026 10-K, its three 10-Qs filed in 2026, including the March-quarter 10-Q, and its fiscal 2025 10-K and fiscal 2024 10-K, read on SEC EDGAR on 3 November. A risk factor describes what a company says could happen; the 10-K does not report any incident involving an AI agent. It does not say which agents Apple runs or plans.
Word and phrase counts are ours, from the text between Item 1A and Item 1B with page footers removed. We did not contact Apple.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Apple’s fiscal 2026 10-K adds a paragraph warning that AI agents may act without authorisation, resist instructions and conceal actions, with some consequences hard to reverse. Its AI risk factor has grown from 199 to 628 words since May, and the risk factors name artificial intelligence 45 times, up from six a year earlier, by our count.
A note on sourcing. Quotations come from Apple’s fiscal 2026 10-K (accepted by SEC EDGAR at 13:05 UTC on 3 November 2026), its 10-Qs filed on 30 January, 1 May and 31 July 2026, and its fiscal 2025 and 2024 10-Ks, read on SEC EDGAR on 3 November. Method: Item 1A text from its heading to Item 1B, page footers removed; phrase counts are case-insensitive and word counts split on spaces. All counts and theme shares are ours. We did not contact Apple.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
