AIHere

Calif.io shows AI-assisted untethered jailbreak for iOS 27

At Hexacon in Paris, the Calif.io team demonstrated an untethered jailbreak working up to iOS 27.0.1, including on the iPhone 18 Pro, with AI helping find the exploit chain.

AIHere News Desk· 2 min read🇺🇸 United States🇫🇷 France
A smartphone with an unlocking padlock icon surrounded by an abstract AI figure scanning lines of code.
Photo: MIKI Yoshihito / Wikimedia Commons, CC BY 2.0

Security research team Calif.io has demonstrated the first jailbreak for iOS 27, presenting it live on stage at the Hexacon security conference in Paris. According to details reported on October 16, 2026, the jailbreak works up to iOS 27.0.1, Apple's latest public iPhone release, and runs on the newest iPhone 18 Pro family devices.

The jailbreak is described as untethered, meaning it survives a reboot. The phone does not need to be connected to a computer to reapply the jailbreak after each restart, a capability that has become rare for current iOS versions.

AI behind the exploit chain

Calif.io has not yet disclosed which vulnerabilities it used. What is known is that artificial intelligence helped the team discover the exploit chain. Hexacon, which focuses on offensive security, has framed this year's edition explicitly as taking place in the "Mythos era", and its organisers promoted the session as a brand-new iPhone jailbreak shown on stage.

The group drew attention in August when it used AI to turn an Apple patch into a working exploit for a serious screen-sharing flaw. It is possible, though unconfirmed, that a similar technique was used here, for example by comparing current iOS 27.2 betas against iOS 27.0.1 to locate security fixes. Because Apple has not documented any such fixes yet, that kind of search would likely consume a large number of tokens.

A technique attackers can copy

Building exploits for already-patched bugs by analysing update packages is spreading quickly. In the screen-sharing case, another exploit, apparently not from Calif.io, circulated soon after the fix. The most capable security-focused language models are currently restricted to selected companies and researchers, but open-weight models carry no such limits, raising concern that attackers with the right prompts could produce similar tools.

Apple has not yet responded to the jailbreak. Once the specific vulnerabilities become clear, the company could ship a fix in a software update. Calif.io's X account has so far offered no further details.

#AI#cybersecurity#iOS 27#jailbreak#Apple#Hexacon

Comments

Be respectful. Comments are checked automatically; spam, insults and personal data are removed. Questions may be answered by AI.

No comments yet. Start the discussion or ask a question.

Related