Calif's iPhone hack: an iOS 27.0.1 jailbreak shown at Hexacon, what Apple has patched, and what role AI played

Calif's iPhone hack, shown at Hexacon in Paris on October 16, 2026, was a reboot-persistent jailbreak of an iPhone 18 Pro on iOS 27.0.1, plus Signal and iMessage demos. Calif hasn't published the bugs or said whether AI found them. What's confirmed, what Apple has patched, and what's only reported.

A glass phone showing a root shell inside a circular restart arrow, with a lit open padlock in the arrow's gap, on amber, titled Survives a restart

What’s new, in 20 seconds

  1. On October 16, 2026, security firm Calif showed four iPhone demos at the Hexacon conference in Paris, including a jailbreak that survives a reboot and one-click Signal and iMessage attacks. Its post says only that 'everything will be released in due time'; it names no bugs, CVEs or AI tools.
  2. Video of the demo shows an iPhone 18 Pro on iOS 27.0.1, build 24A446, according to Decrypt and iClarified. Apple's security releases page lists no CVE fixes in iOS 27.0.1 (September 28) or in iOS 27.1 for iPhone Duo (October 23), and no newer iPhone update as of October 26.
  3. Apple's iOS 27 notes of September 14 credit Calif for four bugs, two 'in collaboration with Claude and Anthropic Research', but nothing ties them to the jailbreak. Calif has used Claude, Claude Mythos Preview and OpenAI's GPT-5.6 Sol in earlier work; it hasn't said what AI did here.
Contents

Calif's iPhone hack is a set of four demos the security firm showed at the Hexacon conference in Paris on October 16, 2026, led by a jailbreak of an iPhone 18 Pro on iOS 27.0.1 that survives a restart. Calif's post about it, "Today, we hacked what we'd set out to hack: the iPhone," drew more than 1,000 points on Hacker News on October 23, the day Calif said it had also jailbroken Apple's new iPhone Duo. This security story separates what Calif has published from what's been reported. Calif is one of the firms credited, with OpenAI and Wiz, with finding the PHP2Shell flaws.

This story is based on Calif's October 16 post and its earlier research posts, Apple's security releases page and iOS 27 security notes, and reporting by Decrypt, iClarified, GBHackers and Redmond Pie, checked October 26, 2026.

The short version

  • What Calif showed: four videos: a one-click Signal attack, an iMessage demo, a jailbroken phone and a demo it calls "APT," for "Apple Persistent Threat."
  • What it hacked: an iPhone 18 Pro on iOS 27.0.1, according to reporting on the demo video. Calif's post doesn't name the model or version.
  • What it published: no bugs, CVEs or technical details. "Everything will be released in due time," the post says.
  • Apple's patches: none of Apple's published security notes is tied to the demos. iOS 27.0.1 is still the newest update for most iPhones.
  • AI's role: not stated by Calif for this work. Heise reported that AI helped, according to GBHackers; Calif's earlier Apple research used Anthropic's models.

What did Calif actually hack?

Calif's post says the demos were "revealed at Hexacon in Paris today" and embeds four captioned videos. It doesn't say whether any attack is zero-click, which components are involved, or what an attacker could reach.

The device details come from the videos. Decrypt reported on October 16 that a clip posted by researcher Billy Ellis shows the phone rebooted, unlocked with its normal passcode and then opening a root shell, with the terminal showing iOS 27.0.1 and about two minutes of uptime. iClarified reported on October 17 that the output identifies the hardware as iPhone19,2 and the build as 24A446, and noted the demo doesn't bypass the passcode. Decrypt called it "the first time an untethered, permanent jailbreak is successful in such devices." Redmond Pie reported that Bruce Dang and Thai Duong gave the Hexacon keynote.

On October 23, iClarified reported that Calif had posted a six-second clip on X showing a jailbroken iPhone Duo on its launch day. The clip doesn't show a restart or name the software build, so it isn't clear whether the same chain was used. Apple's security releases page lists iOS 27.1 for iPhone Duo, released October 23.

There's no public jailbreak tool. iClarified reports that Calif researcher Stefan Esser told Heise the company "does not plan to release a tool for end users, although it may publish a proof of concept."

Has Apple patched the bugs Calif used?

Not in any release Apple has documented. Apple's security releases page, checked October 26, lists iOS 27.0.1 (September 28) and iOS 27.1 for iPhone Duo (October 23), both without published CVE entries, and no newer update for other iPhones.

ReleaseDateDevicesPublished security content
iOS 27September 14, 2026iPhone 11 and laterYes; credits Calif for four bugs
iOS 27.0.1September 28, 2026iPhone 11 and laterNo CVE entries
iOS 27.1October 23, 2026iPhone Duo onlyNo CVE entries

Table 1No Apple release since the Hexacon demo documents a fix tied to it. Source: Apple security releases page and iOS 27 security notes, checked October 26, 2026.

Last verified: October 26, 2026

Calif says it reports bugs to Apple. In a July 27 post, it said it had reported 38 vulnerabilities to Apple since March. For the iOS 27.0.1 chain, iClarified relays Esser telling Heise that the bugs "had already been reported to Apple or were due to be disclosed soon," and that "most were already fixed in the iOS 27.2 beta." Apple hadn't commented publicly on the demo, Decrypt reported. Heise's original report wasn't available to 9to5AI, so these statements are reported, not confirmed.

What role did AI play?

Calif hasn't said. Its Hexacon post doesn't mention AI, and nothing Calif has published links the iPhone chain to a model. GBHackers, citing Heise, reports that AI helped find the exploit chain, without detail.

What is on the record is Calif's broader use of AI against Apple software:

  • April 15, 2026: Thai Duong wrote "Learning to Jailbreak an iPhone with Claude (Part 1)," an analysis of an older leaked exploit kit for iOS 17.2.1. He estimated 40 to 80 hours, much of it checking Claude's explanations, and advised: "let the AI assist, not lead."
  • May 14, 2026: Calif said Anthropic's Claude Mythos Preview "helped identify the bugs and assisted throughout exploit development" for a macOS kernel exploit on M5 hardware, with human experts needed to get past Apple's Memory Integrity Enforcement.
  • September 14, 2026: Apple's iOS 27 notes credit "Calif.io in collaboration with Claude and Anthropic Research" for an AVEVideoEncoder bug (CVE-2026-65410) and Bruce Dang of Calif with the same collaboration for a Foundation bug (CVE-2026-65409). Two more Calif credits, in TCC and the kernel, name no AI. Decrypt notes that nothing in the notes ties these bugs to the jailbreak.
  • October 21, 2026: for PHP2Shell, OpenAI says Calif used GPT-5.6 Sol, as our PHP2Shell story details.

The same iOS 27 notes also credit OpenAI for a kernel bug and "OpenAI Codex Security" in their additional recognition. AI labs have become a visible source of Apple bug reports, part of the wider shift examined in our analysis of OpenAI's Patch the Planet framework. Anthropic's own program for open-source code is covered in our OSS Scanner story.

What should iPhone owners do?

Install updates as Apple releases them. There's no public jailbreak and no published bug detail, so there's nothing specific to mitigate. Hacker News commenters noted the usual trade-off: staying on an old iOS version to keep a jailbreak means missing security fixes.

What's confirmed, reported and uncertain

  • Confirmed (Calif): the four Hexacon demos on October 16, the "released in due time" statement, the earlier AI-assisted research above, and the 38 reports to Apple since March.
  • Confirmed (Apple): the release dates in the table, the September 14 credits to Calif with Claude and Anthropic Research, and no documented fix tied to the demos.
  • Reported: the iPhone 18 Pro, iOS 27.0.1 and build 24A446 details (Decrypt, iClarified, from the video); the iPhone Duo clip (iClarified); Esser's statements on disclosure, the iOS 27.2 beta and a possible proof of concept (Heise, via iClarified); AI helping find the chain (Heise, via GBHackers).
  • Uncertain: which bugs the chain uses, whether AI found any of them, whether the Signal and iMessage demos are zero-click, and when Apple will ship fixes.

Read next

Comments (0)

More on Security

View all
Agents · NewsThe Week in AI #5: AI-found PHP bugs hit the wild, Anthropic's IPO numbers leak, and Meta's $825 billion in commitmentsAI news for October 19–25, 2026: PHP2Shell and OpenAI's Patch the Planet framework, Anthropic's reported $80.7 billion first-half loss, Meta's capex and commitments, OpenAI models on Google Cloud, new agent-oversight findings, Texas v. OpenAI and DeepSeek's reported $15 billion round.·7 min read
Agents · AnalysisAI-found vulnerabilities and maintainers: Sam Altman says we can't 'just patch all the bugs,' Cloudflare's CEO expects weekly alarms, and the PHP Foundation says it's under-fundedAI-found vulnerabilities are arriving faster than maintainers can fix them. What Sam Altman, Cloudflare's Matthew Prince and the PHP Foundation's Elizabeth Barron said between October 21 and 24, 2026, what the numbers show, and who funds the people who write the patches.·9 min read
Hardware · NewsUnitree Go2 vulnerability CVE-2026-115995: a critical flaw in the robot dog's AI chat module, which firmware is affected, and the 1.1.11.5 fixThe Unitree Go2 vulnerability CVE-2026-115995, published October 24, 2026, is a command injection flaw in the robot dog's AI voice-chat service, rated critical. The CVE record lists Go2 firmware 1.1.11.0 to 1.1.11.4 as affected and 1.1.11.5 as fixed. What's known, what isn't, and what owners can check.·5 min read
Developers · GuideClaude auto mode is now the default: what it does, how it differs from accept edits and bypass permissions, and how to turn it offClaude auto mode has been Claude Code's default since August 14, 2026, and on October 22 Anthropic said Claude's web, desktop and mobile apps will follow. How the classifier works, how it compares with accept edits and bypass permissions, and how to switch back to manual.·9 min read
The Week in AI · every Friday

Get the cluster, not just the headline.

The week’s biggest stories in AI and tech, what they mean, and what to read next.

Free. One email a week. Unsubscribe in one click.