OpenAI has expanded ChatGPT’s capabilities on Mac computers with a new feature that allows the AI chatbot to interact directly with Apple’s Messages app, a move that could make everyday tasks more convenient while raising fresh questions about digital privacy. The feature, introduced on Thursday, allows ChatGPT to read, write and send messages, as well as search conversations and generate summaries from a user’s message history.
The integration is likely to attract attention because Apple has long positioned privacy and protection of personal information as key parts of its brand. Giving an AI system access to private conversations could therefore create a new area of concern for users, particularly those who regularly exchange sensitive personal or business information through Messages. Apple had not immediately commented on the development at the time of the report.
OpenAI, however, said the feature only works after users give permission and does not create an index of all their messages. The tool operates locally on a Mac and uses existing macOS capabilities, including AppleScript and Accessibility tools, to interact with the Messages application. Users must approve several permissions during setup, including access to message history, contacts, automation tools and Full Disk Access through Mac System Settings.
The development also places ChatGPT closer to an area Apple is itself targeting with its next generation of AI-powered Siri. Apple is expected to give Siri greater ability to search messages and other personal information stored on users’ devices. While ChatGPT already works with Mac applications such as Xcode, Notes and Terminal, access to Messages is particularly sensitive because it involves highly personal conversations and private communications.
The privacy concerns come as regulators and cybersecurity agencies continue to warn about potential risks associated with AI systems. In December 2025, Nigeria’s National Information Technology Development Agency, through its Computer Emergency Readiness and Response Team, warned about vulnerabilities affecting OpenAI models that could be exploited through indirect prompt injection. The agency cautioned that malicious instructions hidden in webpages or other content could potentially manipulate AI systems, expose information or influence their behaviour. With ChatGPT now moving deeper into users’ personal data and applications, the latest Messages integration is likely to keep the debate over AI convenience versus privacy firmly in the spotlight.
source: nairametrics
