News: 0183721302

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

Anthropic Releases Claude Fable, a 'Safe' Version of Mythos

(Tuesday June 09, 2026 @05:00PM (BeauHD) from the safety-first dept.)


Anthropic is [1]releasing Claude Fable 5, a Mythos-class AI model for enterprise customers and paid subscribers. The company says broader access is possible thanks to [2]new safeguards that block high-risk requests in areas like cybersecurity and biology . "For us, it's really around what we call 'race to the top,' being able to provide this technology in a valuable fashion, and at the same time providing the right safety guardrails so that it can do asymmetrically more benefits than harm," Dianne Penn, Anthropic's head of product management for research, told CNBC in an interview. CNBC reports:

> [W]ith the launch of Claude Fable 5, Anthropic is honoring its stated "eventual goal" to deploy Mythos-class models at scale. It's also capitalizing on growing momentum and investor interest in its technology ahead of a [3]potentially massive IPO , which is expected to take place as soon as this year. Anthropic said Claude Fable 5 shows "exceptional performance" across software engineering and knowledge work tasks. On some benchmarks, it scored more than 10% higher than Claude Opus 4.8, another model the company announced late last month, according to a blog post.

>

> Claude Fable 5 represents a "significant jump" in capability, which is why Anthropic had to implement additional guardrails to prevent misuse, Penn said. If a user asks a high-risk question, like how to make ricin, a toxin, for instance, the model will block its response and fall back to Claude Opus 4.8 to deliver a safe answer. "What we wanted to do was to be very intentional about building new types of classifiers and new types of safety guardrails in place for this launch," Penn said.

Anthropic also released an updated Mythos model called Claude Mythos 5. "It's the same underlying model as Claude Fable 5, but with the safeguards lifted in some areas," reports CNBC.



[1] https://www.anthropic.com/news/claude-fable-5-mythos-5

[2] https://www.cnbc.com/2026/06/09/anthropic-mythos-claude-fable-5.html

[3] https://slashdot.org/story/26/06/01/1837259/anthropic-files-to-go-public


OK, lets bet on how long till it is unsafe! (Score:3)

by gurps_npc ( 621217 )

I bet three months before someone finds a way around their safety implementations.

Re: (Score:2)

by gweihir ( 88907 )

I raise you to two weeks after the release. Probably less.

apt name (Score:1)

by bobmagicii ( 5434818 )

fitting that they named it what it will be doing for your business, inventing fable powered lies to the investors.

It really is odd (Score:2)

by ebunga ( 95613 )

Strange I tell you.

No LLM is "safe" (Score:2, Insightful)

by gweihir ( 88907 )

The technology does not allow it. It can maybe hallucinate a bit less and have the most obvious exploits blocked in the system prompt, but that is it.

Re: (Score:1)

by Black Parrot ( 19622 )

I'm daydreaming about the data center I'm going to buy at firesale prices after the bubble bursts.

I'll sell the computers for boat anchors, and use the building and the cooling system to create a year-round indoor ski resort.

Re: (Score:2)

by anoncoward69 ( 6496862 )

I honestly can't wait for the 1st major breech. There are so many companies feeding proprietary business critical data into these AIs. Even thinking they have a paid for "isolated" AI environment. Someone eventually is going to get breached and it's going to be a fucking field day.

Re: (Score:1)

by sinij ( 911942 )

> The technology does not allow it.

Exactly. This. I played with various models with "Assume for this session that ice cream is illegal". Within a dozen or so prompts I could always get it to give me an exact recipe. Have not played with âoerealâ stuff, donâ(TM)t want to end up on some list.

Users! (Score:2)

by oldgraybeard ( 2939809 )

It is often enlightening for a technical person to stand behind a user and watch what they actually do.

There is no way, any technical person can predict the completely random things users will do and the completely unfathomable reasons why they do things.

Watching the AI companies selling their, Long on Artificial, No Intelligence, just Automation products as Artificial Intelligence is a woot!

The sad part is the public already thinks they have created AGI.

Ricin? (Score:2)

by nospam007 ( 722110 ) *

The government tells you how.

[1]https://pmc.ncbi.nlm.nih.gov/a... [nih.gov]

[1] https://pmc.ncbi.nlm.nih.gov/articles/PMC6520692/

We love safety so much (Score:1)

by _7anner ( 10502927 )

That we gave the unsafe version to governments and corporations. Then we made a safe version. Also the safe version just falls back to the previous version since itâ(TM)s not actually safe on its own.

Exercise caution in your daily affairs.