Anthropic says sorry to users of the new Claude Fable 5 model; says: We apologize for not getting the balance right
Anthropic has revised a safeguard policy for its newly released Claude Fable 5 AI model. This move comes after criticism from researchers and developers over restrictions that were not clearly disclosed to users. The company acknowledged that its approach was flawed and said it would make future refusals and model fallbacks visible. The change comes days after Anthropic launched Claude Fable 5, a public version of its Mythos AI model that includes additional safety measures to prevent misuse in areas such as cybersecurity, biology, and chemistry.In a statement shared with Business Insider, an Anthropic spokesperson said, “We’re changing Fable 5’s safeguards for frontier LLM development to make them visible. Starting this week, flagged requests will visibly fall back to Opus 4.8. On the API, any flagged requests will return a reason for their refusal. We made the wrong tradeoff, and we apologise for not getting the balance right.”
What has Anthropic changed in Claude Fable 5
When Claude Fable 5 launched earlier this week, Anthropic said it had implemented a series of safeguards designed to reduce the risk of the model being used for harmful purposes. Certain prompts related to cybersecurity, biology, and chemistry were automatically routed to less capable models.The company had also disclosed that requests related to AI model development could result in reduced performance without notifying users. According to reports, some developers viewed the measure as a way to limit the model’s use in creating competing AI systems.Under the updated policy, users will now be informed when a request is refused or redirected to another model. API users will also receive a specific explanation when a request is blocked. Anthropic said the safeguards were introduced to address national security risks and prevent “foreign adversaries” from gaining an advantage in the development of advanced chips and large language models.The company further noted that the restrictions affect only a limited category of requests and that the vast majority of coding and machine learning tasks remain unaffected.Claude Fable 5 is based on Anthropic’s Mythos model, which was announced in April and is being made available only to a limited group of government and approved users. Anthropic and external researchers have previously warned that advanced AI systems could potentially be misused in areas such as cyberattacks, biological research, and chemical weapons development.
