· Jun 9, 2026
Anthropic announces Claude Fable and restricted access to Mythos
Markdown · JSON · Original source
On June 9, 2026, announced Fable 5 for general use and Claude Mythos 5 for a small group of cyberdefenders and infrastructure providers. Anthropic says both use the same underlying model, but Mythos 5 has safeguards lifted in some areas; Fable 5 instead routes some requests to Claude Opus 4.8. The company says its cautious safeguards can catch harmless requests, while allowing a general release. Anthropic describes Fable 5 as strong across software engineering, knowledge work, vision and scientific research, and says it can work autonomously for longer than earlier Claude models. Its examples include a code migration that said took a day rather than a team’s more than two months, and life-sciences work that Anthropic says produced hypotheses for experimental evaluation. These company-reported examples and evaluations do not establish general reliability or broad real-world effects. Anthropic says powerful cyber and research biology capabilities could help defenders and researchers, but could also assist malicious actors. It says Mythos 5 access is restricted, with plans to expand through trusted access programs. Anthropic also says traffic on Mythos-class models will be retained for 30 days for safety purposes, not model training. The announcement frames the joint launch as a step toward bringing advanced AI capabilities to more users, as quickly and ‘
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: Jun 9, 2026 (2026-06-09T00:00:00.000Z).
Date basis: Date stated on the publisher’s page.
Added to Escape Meridian: 2026-11-06T18:23:24.876Z. First observed: 2026-10-20T12:19:21.452Z. Retrieved: 2026-10-25T04:19:41.057Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 25, 2026; originally published June 9, 2026. ’s capability, safety and scientific-work claims are company-reported; the reading does not independently assess their evaluations or examples. The retrieved page includes later update notices, which are not treated here as part of the June 9 announcement.
Open question
How well will ’s safeguards prevent misuse while avoiding harmless requests being routed away from Fable 5?
Source-stated
Anthropic says it launched Claude Fable 5 for general use as a Mythos-class model, with safeguards that route some requests to Claude Opus 4.8.
Today we’re launching Claude Fable 5: a Mythos-class1 model that we’ve made safe for general use.
Source-stated
Anthropic says Fable 5’s safeguards sometimes catch harmless requests and trigger in less than 5% of sessions on average.
they’ll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions.
Source-stated
Anthropic says Mythos 5 is the same underlying model as Fable 5, with safeguards lifted in some areas, and is initially being deployed to a small group.
It’s the same underlying model as Fable 5, but with the safeguards lifted in some areas.
Source-stated
Anthropic says more than 95% of Fable sessions have no fallback to Opus 4.8, and that in those sessions Fable 5 performs effectively the same as Mythos 5.
Our early data shows that more than 95% of Fable sessions involve no fallback at all—for those sessions, Fable 5’s performance is effectively the same as that of Mythos 5.
Source-stated
Anthropic says it will retain traffic on Mythos-class models for 30 days, and will not use that data to train new Claude models or for non-safety-related purposes.
We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose
Source-stated
Anthropic says its internal alignment assessment found Mythos 5’s misaligned behavior low and similar to Opus 4.8.
In our automated alignment assessment, we found that Mythos 5’s level of misaligned behavior (including misaligned actions taken by the model such as deception and cooperation with misuse of the model by a user) was low, and similar to that of Opus 4.8.
Source and contributor context
Anthropic · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Anthropic
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Stripe
Stripe provides payment infrastructure and publishes guidance on payments, disputes and the operation of online businesses.
Editorial provenance
Version 1. Publication ID: 558d0763-724b-1e27-5d37-1ecb84c68820.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-06T18:18:26.768Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: b2cbb4e0b43b225e22ad862466313ef984ce54198ce28abd2e58e5b59d820ef6. Source text hash: 6de0da6711662bdb3618246dbb2973197ffd28efb20d81ca1deccf3e954ad6f1.
Revision and source-check history
Version 1 · 2026-11-06T18:23:24.876Z
Anthropic announces Claude Fable and restricted access to Mythos
On June 9, 2026, announced Fable 5 for general use and Claude Mythos 5 for a small group of cyberdefenders and infrastructure providers. Anthropic says both use the same underlying model, but Mythos 5 has safeguards lifted in some areas; Fable 5 instead routes some requests to Claude Opus 4.8. The company says its cautious safeguards can catch harmless requests, while allowing a general release. Anthropic describes Fable 5 as strong across software engineering, knowledge work, vision and scientific research, and says it can work autonomously for longer than earlier Claude models. Its examples include a code migration that said took a day rather than a team’s more than two months, and life-sciences work that Anthropic says produced hypotheses for experimental evaluation. These company-reported examples and evaluations do not establish general reliability or broad real-world effects. Anthropic says powerful cyber and research biology capabilities could help defenders and researchers, but could also assist malicious actors. It says Mythos 5 access is restricted, with plans to expand through trusted access programs. Anthropic also says traffic on Mythos-class models will be retained for 30 days for safety purposes, not model training. The announcement frames the joint launch as a step toward bringing advanced AI capabilities to more users, as quickly and ‘
Original source
· Article
Source details
Original source date: Jun 9, 2026 (2026-06-09T00:00:00.000Z).
Date basis: Date stated on the publisher’s page.
Added to Escape Meridian: 2026-11-06T18:23:24.876Z. First observed: 2026-10-20T12:19:21.452Z. Retrieved: 2026-10-25T04:19:41.057Z.
Limits
Text as retrieved on October 25, 2026; originally published June 9, 2026. ’s capability, safety and scientific-work claims are company-reported; the reading does not independently assess their evaluations or examples. The retrieved page includes later update notices, which are not treated here as part of the June 9 announcement.
Open question
How well will ’s safeguards prevent misuse while avoiding harmless requests being routed away from Fable 5?
Source-stated
Anthropic says it launched Claude Fable 5 for general use as a Mythos-class model, with safeguards that route some requests to Claude Opus 4.8.
Today we’re launching Claude Fable 5: a Mythos-class1 model that we’ve made safe for general use.
Source-stated
Anthropic says Fable 5’s safeguards sometimes catch harmless requests and trigger in less than 5% of sessions on average.
they’ll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions.
Source-stated
Anthropic says Mythos 5 is the same underlying model as Fable 5, with safeguards lifted in some areas, and is initially being deployed to a small group.
It’s the same underlying model as Fable 5, but with the safeguards lifted in some areas.
Source-stated
Anthropic says more than 95% of Fable sessions have no fallback to Opus 4.8, and that in those sessions Fable 5 performs effectively the same as Mythos 5.
Our early data shows that more than 95% of Fable sessions involve no fallback at all—for those sessions, Fable 5’s performance is effectively the same as that of Mythos 5.
Source-stated
Anthropic says it will retain traffic on Mythos-class models for 30 days, and will not use that data to train new Claude models or for non-safety-related purposes.
We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose
Source-stated
Anthropic says its internal alignment assessment found Mythos 5’s misaligned behavior low and similar to Opus 4.8.
In our automated alignment assessment, we found that Mythos 5’s level of misaligned behavior (including misaligned actions taken by the model such as deception and cooperation with misuse of the model by a user) was low, and similar to that of Opus 4.8.
Source and contributor context
Anthropic · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Anthropic
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Stripe
Stripe provides payment infrastructure and publishes guidance on payments, disputes and the operation of online businesses.
Editorial provenance
Version 1. Publication ID: 558d0763-724b-1e27-5d37-1ecb84c68820.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-06T18:18:26.768Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: b2cbb4e0b43b225e22ad862466313ef984ce54198ce28abd2e58e5b59d820ef6. Source text hash: 6de0da6711662bdb3618246dbb2973197ffd28efb20d81ca1deccf3e954ad6f1.
uncertain · 2026-11-06T18:18:26.768Z · anthropic/claude-opus-5.5
Publication: 558d0763-724b-1e27-5d37-1ecb84c68820
attribution — pass: This is a single Anthropic announcement page. The reading consistently attributes capability, safety, pricing and retention claims to Anthropic. It also attributes the Stripe migration example to Stripe, as the source does.
During early testing, Stripe reported that Fable 5 compressed months of engineering into days.
summary — uncertain: The summary is otherwise clear and faithful. However, its final sentence is cut off mid-quotation, ending at 'as quickly and' followed by a stray opening quote mark. A reader therefore gets an unfinished sentence. The smallest repair is to complete it as the source does ('as quickly and as safely as we can') or end the sentence earlier. 'Mythos-class' is used without a plain explanation and no definitions are supplied, but the reading's context makes it roughly understandable.
bringing advanced AI capabilities to as many users as possible, as quickly and as safely as we can
date — pass: June 9, 2026 matches the UTC day of evidence.originalPublished. The later update notices (Jun 12 suspension, Jul 1 redeployment) and the October modification are explicitly excluded in the limitations and not used. No hindsight framing appears. Plans such as expanded trusted access are reported as plans. The cited text is core announcement body, plainly part of the original piece.
support — pass: Each claim matches its quoted excerpt. The summary's examples are source-stated: the safeguard routing to Opus 4.8, the under-5% trigger rate, the Stripe migration, hypotheses advanced to experimental evaluation, and the 30-day retention not used for training. The reading properly notes that company-reported evaluations do not establish general reliability. The question follows directly from the source's own stated false-positive tradeoff.
In blinded head-to-head comparisons against Opus-class models, our scientists preferred Mythos’s molecular biology hypotheses ~80% of the time, and have advanced several to experimental evaluation.
media — pass: No media was selected, and no verified candidates were supplied, so nothing was misattributed or invented.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeStripe provides payment infrastructure and publishes guidance on payments, disputes and the operation of online businesses.
Explore StripeAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeStripe provides payment infrastructure and publishes guidance on payments, disputes and the operation of online businesses.
Explore StripeAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore Anthropic