The team behind GrapheneOS, a security-focused operating system for Android phones, sharply criticized the way Linux is built on X on Sunday evening. “Linux was created as a rejection of robust software architecture and tooling,” GrapheneOS wrote.
According to them, it is a monolithic kernel without internal isolation, written in memory-unsafe code, so “a typo in an obscure driver for an unused feature can easily create a full privilege code execution vulnerability.”
The posts are part of a discussion about whether artificial intelligence introduces bad code. GrapheneOS responded that models cannot be blamed for this, as according to them, hundreds of serious vulnerabilities that previously existed are discovered and fixed in Linux every month. They added that the fact that bugs are now found so easily “creates immense pressure to finally use better tools and approaches.”
The post gathered over 2,000 likes and reached Hacker News, where it faced criticism.
User yjftsjthsd-h ironically asked if that is why GrapheneOS is based on Minix, a microkernel system, even though GrapheneOS is built on top of Android and its Linux kernel.
User qalmakka argued that in 1991, microkernels were “25 years too early,” because computers at the time did not have enough power for them. User somat, meanwhile, reminded that the same applies to Unix, bringing up Tom Van Vleck’s memory of a lunch with Dennis Ritchie: “We left all that out. If there’s an error, we have a routine called panic, and when it is called, the machine crashes, and you holler down the hall, ‘Hey, reboot it.'”
The grapheneos profile on Hacker News, which writes on behalf of the project, responded with accusations of attacks against the team. Later, yjftsjthsd-h wrote that the technical part of the critique is essentially correct, but not the claim about “rejection.”
This is not the first such debate for the project, as last December GrapheneOS also criticized the privacy claims of Sailfish OS.
The debate comes at a time when artificial intelligence is finding more and more bugs in Linux. Last week, Microsoft announced that its FORGE lab, using the MDASH tool, scanned a portion of the kernel and submitted 59 public bug reports, which have so far resulted in 33 CVE identifiers and 38 accepted fixes. One of the bugs had existed for at least 21 years.
Linus Torvalds has already spoken of a “new normal” because of these tools, and one maintainer stepped back from some reviews due to patches written by AI.
However, Microsoft admits that they scanned only a small part of the kernel and concludes that more “tokens” might yield more found bugs, but not a secure system—which is actually close to the point made by GrapheneOS.


