
What happened
Researchers at the Objective-See Foundation found a now-patched bug in the macOS ChatGPT app that could take over ChatGPT on a victim's computer, exposing chat logs, browser sessions and other stored data. OpenAI logged the flaw and fix on September 25.
Why it matters
The bug shows how much system access AI apps like ChatGPT get to work at all, so a small security slip there could expose a user's chats and connected sessions, not just one feature.
What to watch
Patrick Wardle says he has submitted another ChatGPT-related finding, tied to OpenAI's Dots assistant, and OpenAI is reviewing it; he presents his AI macOS bug analysis at Objective by the Sea in November.
WHO IT HITSPeople who use the macOS ChatGPT app on their work machines, and the security and IT teams who approve and manage such AI tools on employee laptops, are the ones most directly affected, since the flaw could have exposed both chat logs and browser sessions on a device.
Summaries like this, every morning.
The finding came from the Objective-See Foundation, a group focused on Apple security, and its analyst Patrick Wardle says the vulnerability was "insanely trivial" to exploit. The design of the macOS ChatGPT app depends on digital signatures checked at three layers of remove from a request, so that a component can only act on behalf of trusted OpenAI software. Wardle's team showed that a trusted script interpreter could accept an untrusted script, and that spawning it three times let the malicious script satisfy those validity checks and reach the main ChatGPT process, which is why the gap mattered out of proportion to its size.
Wardle has been looking at several AI macOS applications, not only ChatGPT. He recently found a flaw, now patched, in the dictation feature of Meta's new Muse AI assistant that could have let a local attacker grab a mishandled authentication token and reach user data. He also says he has submitted a new vulnerability finding to OpenAI related to the link between ChatGPT and the company's new always-on Dots AI assistant, which OpenAI is reviewing.
The timing of this disclosure matters because AI assistants are being granted broad reach into everyday machines to function. Wardle's view, that security is often an afterthought as companies rush features out, suggests the test of these platforms may be less about how fast they patch one bug and more about whether security keeps pace as their access grows. That reading is Wardle's, not a measured industry trend.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Oxide Computer Co., which raised $445 million in October to expand rack production, built its business on the…
OpenAI now requires gateways and routers that provide third-party access to its AI models to be reviewed and a…

TikTok's moderation AI in 2026 evaluates behavioral signals such as Session Diversity, Hardware Telemetry and…

A Zenn article by Hiromi lays out Continuous Reasoning Acquisition (CRA), a design for capturing veteran judgm…

Running the official harness, Claude Opus 5.5 wrote an ARC model with no gradient descent and no forward calls…

Major US tech companies are refocusing generative AI competition from answer quality toward agent capability…
