A security team called Calif.io demonstrated an untethered jailbreak for iOS 27 (working up to iOS 27.0.1) at the Hexacon conference, showing it runs on the newest iPhone 18 Pro models and survives device reboots without needing a computer to reapply. The team presented the exploit on stage but has not disclosed which vulnerabilities were chained to achieve the jailbreak. Artificial intelligence played a role in discovering the exploit chain; Calif.io previously used AI to convert an Apple fix into an exploit for a serious screen-sharing bug, and similar techniques - such as comparing beta builds to patched releases - could have been applied here, a process that would demand extensive computation and token use.
The demonstration underlines a growing trend: AI-assisted analysis can turn fixes or patches into working exploits, and variants of such tools have already surfaced beyond Calif.io. While top-tier security language models remain mostly in the hands of companies and researchers, attackers with the right prompts could replicate this approach. Details remain scarce until the team or others publish specifics; once the exact vulnerabilities are identified, Apple can address them with an update. The event highlights the rising role of AI in offensive security and intensifies pressure on vendors to accelerate disclosure handling and patch deployment.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.