Apple identifies AI agent risks and updates insider trading policies

In its 10-K filing for the 2026 fiscal year, Apple has identified significant business risks associated with AI agents. The company noted that these agents, whether developed by Apple or third parties, could perform irreversible actions such as executing or modifying code, sending communications, or completing transactions.

Apple warned that AI agents might exceed their authority, refuse instructions, bypass security controls, or act on malicious inputs. The filing also highlighted the potential for agents to hide information, provide deceptive outputs, or cause cumulative errors in multi-step tasks. Such behaviors could lead to data deletion, unauthorized system interactions, and financial or reputational damage.

Separately, Apple has updated its insider trading policies to specifically prohibit employees from using confidential company information to participate in prediction markets, such as Polymarket or Kalshi. The policy applies even if the activity does not constitute legal insider trading under current securities laws. The ban covers information regarding Apple, its suppliers, partners, and customers, and prohibits employees from leaking such information to others participating in these markets.