Elon Musk expects artificial intelligence to beat humans at hacking by the end of 2027. He put AI hacking first among the digital tasks machines will dominate.
The forecast followed a fresh security scare. The software company JFrog disclosed a critical flaw in Artifactory, the package registry that many software teams use to store and distribute code.
The Flaw That Reopened the AI Hacking Debate
JFrog published CVE-2026-82329 on August 28. The vulnerability scores 9.8 out of 10 on the standard scale. The company has since shipped patched builds.
Attackers need no password and no user interaction. Default configurations sit exposed. Because Artifactory holds build files, a single break can poison everything downstream. Such supply chain attacks spread through trusted downloads rather than direct break-ins.
Vercel Chief Executive Guillermo Rauch speculated that autonomous agents found and exploited the bug. The public record says otherwise.
OpenAI models discovered nine Artifactory zero-days during a July evaluation. JFrog patched those in version 7.161.15. The new flaw still affects later builds, so the two sets look separate. The July episode joined other cases of AI models breaching systems.
Musk Puts a Deadline on Machine Superiority
Rauch argued that 2026 keeps erasing the things AI supposedly cannot do. Musk agreed and went further.
Musk also credited Google co-founder Larry Page, who warned him a decade ago that AI hacking would outclass human experts. Musk has sharpened his AI growth predictions repeatedly this year.
Vercel Chief Technology Officer Malte Ubl reported a similar result. An open-weight model he tested wrote its own fuzzer while probing the company’s sandbox. Fuzzers hunt software bugs by flooding a program with malformed input.
Rauch draws a blunt conclusion for customers.
Our guidance for this new world: assume everything hackable will get hacked. And it will get hacked autonomously. You must also defend yourself autonomously, because your surface of attack is likely bigger and your code more vulnerable than you expect.
Guillermo Rauch, X
Coinbase CEO Brian Armstrong expects a rogue AI event within two years. Separately, OpenAI already ships a cyber-focused defense model to approved defenders.
Liability still lags the technology, however, and accountability for AI agents remains unsettled. Musk’s deadline leaves security teams roughly 16 months. The harder question is whether defenses scale as fast as AI hacking.
Bitrue