Musk Sees AI Hacking Exceeding Human Ability by 2027

By the close of 2027, Elon Musk believes AI will surpass human capabilities in hacking. He ranked hacking as the foremost of the digital tasks AI will come to dominate.

His prediction came after a recent security incident. The software firm JFrog revealed a critical vulnerability in Artifactory, a package registry widely used for code storage and distribution.

The Flaw That Revived the AI Hacking Discussion

On August 28, JFrog disclosed CVE-2026-82329. The flaw carries a severity rating of 9.8 out of 10. Patched versions have since been released.

No password or user interaction is required to exploit it. Default settings remain exposed. Since Artifactory holds build files, one breach can contaminate all downstream software. These supply chain attacks spread via trusted downloads rather than direct intrusions.

Guillermo Rauch, Vercel's CEO, speculated that autonomous agents discovered and exploited the vulnerability. The public record, however, indicates otherwise.

In a July evaluation, OpenAI's models found nine Artifactory zero-days. Those were fixed in version 7.161.15. The latest vulnerability persists in later builds, indicating the two groups are distinct. That July incident added to a growing list of AI-driven system breaches.

Musk Sets a Timeline for Machine Dominance

Rauch argued that 2026 keeps disproving what AI was thought unable to do. Musk concurred and took it a step further.

AI will be able to do anything digital (that doesn't require shaping atoms) at a superhuman level by the end of next year

— Elon Musk (@elonmusk) August 31, 2026

Musk also acknowledged Google co-founder Larry Page, who warned him a decade ago that AI hacking would surpass human expertise. Musk has sharpened his AI growth forecasts multiple times this year.

Malte Ubl, Vercel's CTO, reported a similar outcome. An open-weight model in his tests created its own fuzzer while examining the company's sandbox. Fuzzers detect software bugs by bombarding a program with malformed data.

Rauch draws a stark conclusion for users.

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 incident within two years. Meanwhile, OpenAI has released a cyber defense model to vetted defenders.

Accountability has not kept pace with the technology, and responsibility for AI agents is still unresolved. Musk's timeline gives security teams about 16 months. The more pressing issue is whether defensive measures can keep up with AI-driven attacks.