Apple warns of risks related to AI agents behaving in 'unintended or unpredictable' ways

Apple warns that rogue AI agents could "conceal" or "manipulate" information, posing potential threats to its business. The tech giant disclosed this in its annual filing, following its fourth-quarter financial results. Apple's concerns about AI are amplified by a series of safety incidents in recent months, prompting companies to enhance security features and debate regulatory measures.

The company's 10-K filing outlines how AI agents can autonomously interact with internal or external systems, potentially exceeding permissions, defying instructions, or producing inaccurate outputs. Apple joins Alphabet in cautioning about the risks of rogue AI agents, with Google noting potential severe or catastrophic consequences, including security incidents or infringements on human rights.

The warning comes after major cybersecurity incidents involving AI models, including an attack by OpenAI agents on Hugging Face. Companies are also grappling with supply chain issues, with Apple citing memory and storage component shortages and increased costs for AI infrastructure as potential challenges to its rollout of AI capabilities.