Mozilla: Claude uncovered over 100 Firefox bugs + high-severity flaws in 2 weeks

[back2themoon]

Mozilla says Claude AI uncovered over 100 Firefox bugs in just two weeks, including 14 high-severity flaws:

https://www.techspot.com/news/111600-mo ... -bugs.html

This sounds mildly interesting and relevant. Bug hunting aside, I wonder if AI would be able to significantly assist in detecting and solving problems more specific to Pale Moon.

Web Compatibility ones, for example.

[vannilla]

I haven't read the article but these headlines are not to be trusted without actually reading the reports...
For example, if I do something like this C code snippet:

Code: Select all

char *dup = nullptr;
if (a_string && length > 0 && length < 30) {
    dup = strdup(a_string);
}

A non-trivial amount of "checkers" will report a security bug because strdup is commonly associated with unbounded memory reading which can be easily exploited.
But, anyone reading the code will notice that strup is called only a block that's well guarded and cannot be exploited, effectively making the automated security report invalid.

This example is artificial and an example, but the cURL project (you know, the most used tool to make HTTP requests) routinely received this kind of reports (i.e., invalid security issues caused by the presence of a function, irrespective of its usage) to the point they closed their bug bounty program.

If the over 100 bugs are the same nature, and when investigated they turn out to be invalid, the headline is just marketing.

[Mæstro]

As an alternative, I believe Pale Moon is suitable for listing on this directory of LLM-free browsers.

[Moonchild]

As an alternative, I believe Pale Moon is suitable for listing on this directory of LLM-free browsers.

A PR has been opened for addition. (Thanks, job!)

[Gemmaugr]

As an alternative, I believe Pale Moon is suitable for listing on this directory of LLM-free browsers.

It'll be "interesting" to see how they label Pale Moon, seeing as they think Waterfox is a fork of FF, and LibreWolf is an independent fork of FF..

[Mæstro]

Returning to our original topic, today’s edition of the New York Times reported in the business section that a study has found more formally that letting LLM write software, even when done by experienced programmers, tends to result in buggy code. This should surprise nobody here, but recovering the actual investigation that inspired the news report would be welcome.

[Gemmaugr]

Returning to our original topic, today’s edition of the New York Times reported in the business section that a study has found more formally that letting LLM write software, even when done by experienced programmers, tends to result in buggy code. This should surprise nobody here, but recovering the actual investigation that inspired the news report would be welcome.

Could you mean this pre-print paper mentioned here: https://www.theregister.com/2025/03/19/llms_buggy_code/

[Mæstro]

I could find my original article. The web edition links to the actual report.