A fully updated iPhone 18 Pro hacked on stage in Paris stays compromised after a reboot
On October 16 at the Hexacon conference, the firm Calif showed a persistent jailbreak of iOS 27.0.1. The flaws have not been published and no patch is known. Here's what it means for users.
Translated from French by AI. The original French article is the reference version.

The iPhone reboots. The team unlocks it with the usual passcode, then opens a remote session (SSH) with "root" privileges, the highest level of access on the phone. The terminal shows the device has been running for about two minutes. The scene took place on Friday, October 16, in Paris, on the stage of Hexacon, a conference devoted to offensive security. The video was published by researcher Billy Ellis, according to the site Decrypt.
The demonstration was given by Calif, a security firm. The phone is an iPhone 18 Pro running iOS 27.0.1, the latest public version of iOS. What sets this jailbreak apart is that it is "untethered": it reinstalls itself after a reboot, with no computer plugged in. According to Decrypt, this is the first time such a feat has been achieved on these devices.
Why a jailbreak worries defenders too
For enthusiasts, a jailbreak lifts Apple's restrictions: installing unapproved apps, accessing the entire file system. But a chain of flaws that can take full control of an up-to-date iPhone and survive a reboot looks like what commercial spyware vendors sell, Decrypt points out. As in the case of Candiru and the Al Jazeera journalists, the value of these chains lies in their staying secret.
Yet Calif has published nothing. The flaws used are not named and no tool has been released. The company promises that everything will be revealed "in due course," without giving a date. Its newsletter announces four demonstrations, including one targeting Signal with a single click and another targeting iMessage, with no further details. Apple has not commented: the company never confirms a flaw before releasing a patch. It is also unknown whether Calif has passed its chain on to Apple.
What role does artificial intelligence play?
The German site heise claims that AI "again" helped find the exploit chain, without saying how. What is established concerns Calif's earlier work. In May, the firm said it had built a macOS kernel exploit in less than a week using Claude Mythos Preview, Anthropic's model. The iOS 27 security notes, published on September 14, credit Calif "in collaboration with Claude and Anthropic Research" for two flaws. According to Decrypt, these cause a denial of service or a crash and are not linked to the jailbreak. A comparable race played out on Chrome, hacked in six days with an OpenAI model.
What to do in the meantime?
No attack code is circulating at this stage. So there is no emergency measure specific to this demonstration. The immediate risk lies elsewhere: some sites are already offering fake "iOS 27.0.1 jailbreaks" for download. None of them is genuine.
- Do not install any tool or configuration profile presented as this jailbreak.
- Apply the next iOS update as soon as it is released: Apple generally fixes these chains quickly once the details are known.
- For people at risk (executives, journalists, elected officials), iOS Lockdown Mode reduces the attack surface.
How long iOS 27.0.1 stays vulnerable now depends on a single date: the one Calif chooses for publication.
Method: AI-assisted monitoring, verified sources, article reviewed and approved by Frédéric LENQUETTE. Our editorial policy.
Share this article