News for hackers. Stuff that matters.
Re:Sonnet 5.5 by ghoshbishakh on Monday September 28, 2026 @02:27PM (#49882304)
So sonnet is better than Fable now? That Fable which was too dangerous to release? I am so confused now.
Re:Sonnet 5.5 by pebbly_bread on Monday September 28, 2026 @03:21PM (#49883086)
Mythos is what they thought was too dangerous to release, fable was what they made after they worked on cybersecurity detection. As they say in the notes, this version of sonnet now has a similar screening process
Re:Sonnet 5.5 by rs_rs_rs_rs_rs on Monday September 28, 2026 @04:19PM (#49883788)
Mythos and Fable are the same llm. Fable has an extra tool that's in front of it that decides to accept the promp or not.
Re:Sonnet 5.5 by usef- on Monday September 28, 2026 @07:13PM (#49885767)
And Sonnet has a similar classifier in front according to the article:

> it’s the first Sonnet model to launch with cyber safeguards

Re:Sonnet 5.5 by dbbk on Monday September 28, 2026 @04:31PM (#49883948)
Hence why it is no longer dangerous, yes
Re:Sonnet 5.5 by emil-lp on Monday September 28, 2026 @06:19PM (#49885260)
If you consider security through obscurity a safe route, yes
Re:Sonnet 5.5 by usef- on Monday September 28, 2026 @07:15PM (#49885787)
How is this security through obscurity?

That term is about hiding a system's design in order to secure something, rather than having secure design.

Re:Sonnet 5.5 by emil-lp on Tuesday September 29, 2026 @12:57AM (#49888443)
You answered your own question!
Re:Sonnet 5.5 by usef- on Tuesday September 29, 2026 @01:01AM (#49888468)
This is not relying on a hidden design.
Re:Sonnet 5.5 by emil-lp on Tuesday September 29, 2026 @02:55AM (#49889220)
The design allows users to bypass the security by luck and chance.

A secure system is impenetrable unless you have the key.

Re:Sonnet 5.5 by usef- on Tuesday September 29, 2026 @03:39AM (#49889568)
It sounds like you're thinking of cryptography, not general security.

The world uses many security products that have false positives and false negatives (firewalls, intrusion detections, wafs, fraud detection, spam...). Those aren't generally considered security through obscurity.

They openly talk about the system and its drawbacks here if you're interested: https://www.anthropic.com/news/fable-safeguards-jailbreak-fr...

It's paired with rate limits, monitoring, account control, multiple classifiers, a deliberate safety margin, model design, and other things too.

(I think this was also why they faced the controversy over not having zdr in fable: they wanted to use logs to detect repeated attempts, etc. Possibly a bigger change with Opus/Sonnet is that it's zdr with the classifiers?)

Re:Sonnet 5.5 by stravant on Tuesday September 29, 2026 @05:33AM (#49890444)
That's not how it works.

The reason you see "dumb" refusals that "should clearly be allowed" is that they're using more traditional deterministic methods to deny prompts rather than just relying on the random LLM which you could bypass by luck.

Re:Sonnet 5.5 by mnicky on Monday September 28, 2026 @03:09PM (#49882933)
Well, it's performance "surface" (is there a better term for this?) is probably very narrow compared to Fable :)
Re:Sonnet 5.5 by heyjstn on Monday September 28, 2026 @02:59PM (#49882798)
doom marketing at its finest
Re:Sonnet 5.5 by solenoid0937 on Monday September 28, 2026 @03:57PM (#49883508)
Not really, they never released Mythos. And they never said Fable was dangerous. They've been very consistent
Re:Sonnet 5.5 by Art9681 on Monday September 28, 2026 @06:30PM (#49885368)
You have to actually spend some effort reading the article they published to answer your own question.