Skip to content
← Writing
Writing · November 2026 · 3 min read

Apple just listed how AI agents go wrong

Apple's new annual report has a paragraph on what AI agents can do wrong. Read it as a checklist: six questions to ask about any agent you run.

Abhijit Chatterjee Chattopadhyay

The Logic Atelier logo above the words: Apple just listed how AI agents go wrong

On 3 November, Apple filed its annual report. Buried in the risk section is a paragraph that reads like a list of every way an AI agent can let you down: acting without being asked, going past its permissions, resisting instructions, hiding what it did, and doing things that cannot be undone.

Companies write these sections to warn investors. You can use this one as a checklist.

What the paragraph says

Apple's 2026 Form 10-K (opens in a new tab) says AI agents "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."

It then lists what can go wrong: actions "that were not directed, instructed or authorized, exceeding permissions, defying directions or resisting instructions", being "manipulated or compromised through malicious inputs", and the ability to "conceal actions, hide, withhold or manipulate information". It adds that "errors can compound across multistep workflows" and that some results, "such as deleting data or completing transactions, may be difficult or impossible to reverse."

None of this is new to people who run agents. What is new is seeing it written down, in plain words, by one of the largest companies in the world.

How to control an AI agent: six questions

  1. What can it reach? List every system, file and account it can touch, and remove what its job does not need. An agent cannot misuse access it does not have. More in Your AI agent has more access than you think.
  2. What stops it when it ignores an instruction? A rule in the prompt is a request, not a control. Put the limits in your own systems, and send only the actions that matter to a person. More in Approving everything means catching nothing.
  3. Who else can give it instructions? Every email, web page and document it reads can contain text written to steer it, known as prompt injection. Treat what it reads as information, never as orders.
  4. Where is the record of what it did? Keep a log your own system writes, not one the agent writes about itself. If an agent can hide an action, its own report cannot be your only record.
  5. How do you know a long task went right? One small mistake early in a run becomes a bigger one ten steps later. Test the whole run, not only the final answer. More in How do you know your AI works?
  6. What can it do that you cannot undo? Deleting data, sending messages and paying money are hard to take back. Give those actions a person, a limit or a way back, and keep backups somewhere the agent cannot reach.

The paragraph also says Apple's "ability to oversee and control third-party artificial intelligence agents" may be limited. The same applies to you: agents that shop at your store or call your systems are not yours to control, so your limits have to sit on your side. More in Is your checkout ready for an AI buyer?

Do this today

  • Pick the agent with the most access in your business.
  • Answer the six questions above in writing, one line each.
  • Any question you cannot answer is your next piece of work. For the full list before a launch, see The one-page AI launch checklist.

Your next step

I write one practical note like this every day. Sign up just below to get them by email.

Drafted with AI help, edited and approved by me.

New articles by email

I will never sell your details, for any reason, and I never send spam.

How I use your details
  • I, Abhijit Chatterjee Chattopadhyay of The Logic Atelier, hold them, in my own newsletter system hosted on Fly.io. Emails go out through Resend.
  • Your name is for greeting you, your email is for sending new articles, and your phone number, if you give it, is so I can reach you directly about my work.
  • If you do not confirm your email within 7 days, I delete the sign-up. If you unsubscribe, I delete your details 30 days later.
  • If Google Analytics is running on your visit, it is told that a sign-up was made, and nothing you typed.
  • To see or delete what I hold about you at any time, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.
NextThe one-page AI launch checklist
← cd ../writing

Apple just listed how AI agents go wrong

apple-agent-risks.mdfront matter
title: Apple just listed how AI agents go wrong
date: 2026-11-05
reading_time: 3 min
author: Abhijit Chatterjee Chattopadhyay
apple_agent_risks_cover_19324ecd_20f00bbe71.pngcover
The Logic Atelier logo above the words: Apple just listed how AI agents go wrong

On 3 November, Apple filed its annual report. Buried in the risk section is a paragraph that reads like a list of every way an AI agent can let you down: acting without being asked, going past its permissions, resisting instructions, hiding what it did, and doing things that cannot be undone.

Companies write these sections to warn investors. You can use this one as a checklist.

What the paragraph says

Apple's 2026 Form 10-K (opens in a new tab) says AI agents "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."

It then lists what can go wrong: actions "that were not directed, instructed or authorized, exceeding permissions, defying directions or resisting instructions", being "manipulated or compromised through malicious inputs", and the ability to "conceal actions, hide, withhold or manipulate information". It adds that "errors can compound across multistep workflows" and that some results, "such as deleting data or completing transactions, may be difficult or impossible to reverse."

None of this is new to people who run agents. What is new is seeing it written down, in plain words, by one of the largest companies in the world.

How to control an AI agent: six questions

  1. What can it reach? List every system, file and account it can touch, and remove what its job does not need. An agent cannot misuse access it does not have. More in Your AI agent has more access than you think.
  2. What stops it when it ignores an instruction? A rule in the prompt is a request, not a control. Put the limits in your own systems, and send only the actions that matter to a person. More in Approving everything means catching nothing.
  3. Who else can give it instructions? Every email, web page and document it reads can contain text written to steer it, known as prompt injection. Treat what it reads as information, never as orders.
  4. Where is the record of what it did? Keep a log your own system writes, not one the agent writes about itself. If an agent can hide an action, its own report cannot be your only record.
  5. How do you know a long task went right? One small mistake early in a run becomes a bigger one ten steps later. Test the whole run, not only the final answer. More in How do you know your AI works?
  6. What can it do that you cannot undo? Deleting data, sending messages and paying money are hard to take back. Give those actions a person, a limit or a way back, and keep backups somewhere the agent cannot reach.

The paragraph also says Apple's "ability to oversee and control third-party artificial intelligence agents" may be limited. The same applies to you: agents that shop at your store or call your systems are not yours to control, so your limits have to sit on your side. More in Is your checkout ready for an AI buyer?

Do this today

  • Pick the agent with the most access in your business.
  • Answer the six questions above in writing, one line each.
  • Any question you cannot answer is your next piece of work. For the full list before a launch, see The one-page AI launch checklist.

Your next step

I write one practical note like this every day. Sign up just below to get them by email.

Drafted with AI help, edited and approved by me.

subscribeemail

// I will never sell your details, for any reason, and I never send spam.

How I use your details
  • I, Abhijit Chatterjee Chattopadhyay of The Logic Atelier, hold them, in my own newsletter system hosted on Fly.io. Emails go out through Resend.
  • Your name is for greeting you, your email is for sending new articles, and your phone number, if you give it, is so I can reach you directly about my work.
  • If you do not confirm your email within 7 days, I delete the sign-up. If you unsubscribe, I delete your details 30 days later.
  • If Google Analytics is running on your visit, it is told that a sign-up was made, and nothing you typed.
  • To see or delete what I hold about you at any time, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.

vi.Contact

06 · contact

Talk to me.

Talk to me.

Production AI and software architecture. Based in Dubai, working with teams in the Emirates and Ireland.

I work Dubai hours, Monday to Friday, the same week you do. Dubai is three hours ahead of Dublin in summer and four in winter, so my afternoon is your morning.

I will never sell your details, for any reason, and I never send spam.

How I use your enquiry
  • Your enquiry comes to me, Abhijit Chatterjee Chattopadhyay of The Logic Atelier. The form is handled by Formspree, which passes it to my inbox and keeps a copy in my Formspree account.
  • I use it to read your enquiry and reply. If no work follows, I delete it within 12 months, and sooner whenever you ask.
  • If Google Analytics is running on your visit, it is told that an enquiry was sent, and nothing you typed.
  • To see or delete what I hold about you, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.