Unconfirmed: Paul Walsh, who is building a rival messaging service, alleged on X1 on 17 November that Apple added macOS beta warnings after Meta's Muse launched but stayed silent on OpenAI's ChatGPT Messages integration; Automatica has not confirmed it.
Walsh discloses at the foot of his post that he and a co-founder are building "a new kind of messaging service" and asks readers to weigh his criticism "with this commercial interest in mind." His post is the only source. Apple, Meta and OpenAI have not responded in the material Automatica reviewed, and Automatica has no earlier reporting on this claim.
Walsh writes that Apple "has introduced scary new warnings about the intrusive permissions that Meta's spyware agent Muse requests, while remaining silent when its close partner OpenAI built a ChatGPT integration specifically designed to access private iMessage conversations using the same permissions." He says he published a technical investigation of the ChatGPT integration, that Proton cited it and that Fortune interviewed him, but that Apple declined to comment. Automatica has not reviewed any of those items.
On the software, Walsh says the macOS 27.3 beta introduces "new controls over which files and applications AI agents and other software applications can access." He says Apple is splitting its "Device Control and Data Access" permission into "System & App Protection Override" and "Accessibility." Customers could lock down specific apps or folders, require an ask-every-time prompt, and see an alert when one app is about to control another, he writes.
He describes the change as double-edged: "While these changes give customers more control over their privacy, they also give Apple more control over which developers can access these capabilities and qualify for exceptions." Walsh argues the controls would let Apple decide which permissions competitors receive.
The unequal-treatment claim is a forecast in his own words, not an observation. "I suspect Apple and its close partners will receive more reassuring warnings for the same permissions, while competitors face messages designed to scare customers away," he writes, adding that Apple's warnings "probably won't apply equally to every developer." The post does not reproduce prompt text for Muse or for the ChatGPT integration.
Walsh also calls Muse and the ChatGPT integration "spyware" and argues the data they reach could be exposed to governments. Those are his characterisations and projections. Automatica has not tested either product.
What would confirm or kill the claim is specific. Apple's macOS 27.3 beta release notes or developer documentation would show what the permission changes are and whether exceptions exist. Side-by-side captures of the prompts shown for Muse and for the ChatGPT Messages integration, under the same permissions, would show whether the wording differs. Uniform prompt wording across developers would undercut his suspicion. A statement from Apple, Meta or OpenAI addressing the allegations would also bear on it.
Automatica has not reviewed Apple beta notes, documentation or screenshots of the prompts. Descriptions of the beta in this report are Walsh's alone.