
OpenAI has launched an Apple Messages plugin for ChatGPT on Mac. It can search and summarize texts, but only after users grant access.
Privacy experts warn it could expose conversations of people who never agreed.
The plugin runs locally and only reads when asked.
What happened
OpenAI rolled out an Apple Messages plugin for ChatGPT on Mac last week. It lets users search iMessage, SMS, and RCS conversations, catch up on threads, draft replies, and send messages, all with explicit user consent.
Why it matters
Security expert Paul Walsh calls it 'one of the most dangerous things I have seen in technology' and says it can function like spyware. Even though the plugin runs locally by default and only reads messages when asked, it makes years of conversations searchable by AI, including messages from people who never opted in.
Ask the AI about this article →
The launch of the Messages plugin reflects a broader trend where AI services are seeking access to more sensitive data. Lookout's research of over 420 million apps shows AI-related apps have expanded their permissions and capabilities over the past year. Richardson noted a steady trend of AI services asking for more data access. This integration uses existing macOS capabilities, not a new Apple API, and OpenAI says it runs locally by default, but the privacy debate centers on the consent gap: one person's opt-in can expose others' messages to AI. Walsh argues this breaks the fundamental protections of end-to-end encryption, even if the encryption itself is not compromised. The concern is not about breaking Apple's encryption but about what happens after messages are decrypted on a Mac, where another system can read them. This could create a 'side door' for authorities, as Walsh describes, if conversation data is stored in the cloud, though OpenAI limits this by not uploading entire histories and keeping content local by default. The plugin is part of a broader expansion of AI access, and the tension between utility and privacy is likely to persist as AI agents gain more access to people's digital lives.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
The Linux Foundation announced the contribution of TRACE (Trust, Runtime Attestation and Compliance Evidence)…

A new Stanford research paper warns that AI chatbots and agents may let advertising dollars influence what the…

Researchers from Bytedance Seed developed EdgeBench, a benchmark measuring how AIs improve on tasks over multi…

OpenAI CEO Sam Altman told TIME the company will have an internal system he'd call AGI by the end of 2026

A large-scale study of 17,022 skills from SkillsMP, the largest open-source skill marketplace, found 520 affec…

Meta halted its planned AI-driven layoffs hours before the first round, after internal documents revealed a pl…
