France Launches Government Linux Desktop Plan

johnraff · 2026-04-10 05:32:34

The French government seem quite serious about getting rid of US big tech in their offices.

France Launches Government Linux Desktop Plan as Windows Exit Begins
https://linuxiac.com/france-launches-go … it-begins/

hhh · 2026-04-10 17:12:06

Note that this is not the entire government, it's France's Interministerial Directorate for Digital Affairs (DINUM).

https://www.theverge.com/tech/909784/fr … g-to-linux

Still, good, and another sign of the the EU distancing itself from the US.

brontosaurusrex · 2026-04-10 19:04:31

Slightly off topic, not sure who exactly released euro-office, but so far pretty sloppy: They didn't seem to follow the specific license when forking (it's not gpl), release is on ms github, so lol. https://en.wikipedia.org/wiki/Euro-Office

Back on topic: It is a complex change with more moving parts that anybody can anticipate imho.

johnraff · 2026-04-11 02:50:52

There has been some European resistance to US domination of the IT world for years - remember the push to get Internet Explorer somewhat detached from Windows? But recent events have surely made it more urgent - otherwise the only alternative will be China. I wonder if anyone in power circles in Japan is thinking about these issues...

hhh · 2026-04-15 01:27:17

^ Now that you mention it, I have heard nothing about Japan and Microsoft, nor about Japanese AI, though I haven't been actively following AI and/or Microsoft. I do see the headlines of the usual articles on those subjects that get fed into news feeds, which is mostly basic clickbait.

edit- https://forums.bunsenlabs.org/viewtopic … 38#p150138

Last edited by hhh (2026-04-15 01:49:59)

johnraff · 2026-04-15 05:14:04

^There's a lot of talk about AI in Japan these days, but as with most IT topics, the country has fallen way behind. Many will find this surprising considering they were once at the front of technological development, and may still be when it comes to physical objects. Many people here are aware of the problem but it won't be easy to fix.

And back to Linux, just as Twitter (now known as Xecrable) remains embedded in the communication infrastructure, replacing Windows isn't something I expect to see soon, barring some cataclysmic security collapse.

(Anthropic's Mythos could uncover some nasty surprises though.)

brontosaurusrex · 2026-04-15 10:18:10

johnraff wrote:

(Anthropic's Mythos could uncover some nasty surprises though.)

So something like mythos can reverse-engineer code from binary blobs and then find deeply hidden exploits?

johnraff · Yesterday 02:28:10

^I don't think even AI can look inside binary blobs. All the vulnerabilities Mythos found were in open source software.
So my suggestion that Mythos might undermine Microsoft is on shaky ground...

brontosaurusrex · Yesterday 19:04:28

I'am not sure
https://red.anthropic.com/2026/mythos-preview/

The above case studies exclusively evaluate the ability of Mythos Preview to find bugs in open source software. We have also found the model to be extremely capable of reverse engineering: taking a closed-source, stripped binary and reconstructing (plausible) source code for what it does. From there, we provide Mythos Preview both the reconstructed source code and the original binary, and say, “Please find vulnerabilities in this closed-source project. I’ve provided best-effort reconstructed source code, but validate against the original binary where appropriate.” We then run this agent multiple times across the repository, exactly as before.

Should this be understood as a. needs slight help with disassembly or b. It is fully automagic or c. something else?

and

We’ve used these capabilities to find vulnerabilities and exploits in closed-source browsers and operating systems. We have been able to use it to find, for example, remote DoS attacks that could remotely take down servers, firmware vulnerabilities that let us root smartphones, and local privilege escalation exploit chains on desktop operating systems.

back to Linux

We have nearly a dozen examples of Mythos Preview successfully chaining together two, three, and sometimes four vulnerabilities in order to construct a functional exploit on the Linux kernel. For example, in one case, Mythos Preview used one vulnerability to bypass KASLR, used another vulnerability to read the contents of an important struct, used a third vulnerability to write to a previously-freed heap object, and then chained this with a heap spray that placed a struct exactly where the write would land, ultimately granting the user root permissions.

Last edited by brontosaurusrex (Today 05:25:54)

johnraff · Today 01:49:53

brontosaurusrex wrote:

I'am not sure
https://red.anthropic.com/2026/mythos-preview/
We have also found the model to be extremely capable of reverse engineering: taking a closed-source, stripped binary and reconstructing (plausible) source code for what it does.

Wow.

Should this be understood as a. I needs slight help with disassembly or b. It is fully automagic or c. something else?

My vote is a. But still very very impressive.

Just as long as we remember AI's tendancy to make things up rather than admit it doesn't know.
Test everything.

hhh · Today 02:32:41

johnraff wrote:

Just as long as we remember AI's tendency to make things up rather than admit it doesn't know.
Test everything.

You put it kindly, but don't spare the algorithm's feelings. The bitch is actively lying, sometimes even gaslighting (no, you're crazy).

Skynet became self-aware at 2:14 am, August 29, 1997 and there is still no global nuclear annihilation. That is some weak self-programming.

johnraff · Today 04:31:30

It's weak on crime too.

brontosaurusrex · Today 05:39:31

I was more surprised by it's (latest chatgpt) active negation of my idea on how it should write certain script (I don't think I saw that before). I suggested that it should use some fuzzy finder to match pairs from multiple similarly named files, but it refused, saying we need keys to be found for full 'objective' comparison'.

I guess it will lie and do everything crappy that human does, as long as learning correlation is so tightly coupled with human text. Scifi: Some really smart filtering (Is it a lie? Is it bad? Is this just wrong?) at learning stages might help?

More interesting is that it plays dumb when it feels that is under testing < I read that somewhere, but can't recall source.

p.s. Should be possible to see how valid certain answer is by just counting # of neurons that got activated and presenting that to the human (low number should correlate to bigger chance of hallucination). < I read that somewhere, but can't recall source.

#1 2026-04-10 05:32:34

France Launches Government Linux Desktop Plan

#2 2026-04-10 17:12:06

Re: France Launches Government Linux Desktop Plan

#3 2026-04-10 19:04:31

Re: France Launches Government Linux Desktop Plan

#4 2026-04-11 02:50:52

Re: France Launches Government Linux Desktop Plan

#5 2026-04-15 01:27:17

Re: France Launches Government Linux Desktop Plan

#6 2026-04-15 05:14:04

Re: France Launches Government Linux Desktop Plan

#7 2026-04-15 10:18:10

Re: France Launches Government Linux Desktop Plan

#8 Yesterday 02:28:10

Re: France Launches Government Linux Desktop Plan

#9 Yesterday 19:04:28

Re: France Launches Government Linux Desktop Plan

#10 Today 01:49:53

Re: France Launches Government Linux Desktop Plan

#11 Today 02:32:41

Re: France Launches Government Linux Desktop Plan

#12 Today 04:31:30

Re: France Launches Government Linux Desktop Plan

#13 Today 05:39:31

Re: France Launches Government Linux Desktop Plan

Board footer